Multinomial logistic regression predicting class membership for trajextories of GAD-7, PHQ-9 and MASQ-anhedonia
K L Purves
08 September, 2021
Background
Overview
We perform multinomial logistic regressions using the multinom function from the nnet package to predict latent class membership (most likely trajectory assignment) in our best GMM model for each symptom measure (anxiety, GAD-7; depression, PHQ-9; anhedonia, MASQ CD anhedonia subscale).
Description of best model for each symptom measure
See GMM notebooks and pre-registration for a description of how best models were selected. An overview of the model for each symptom measure and a brief explanation for its selection is described here.
Depression (PHQ-9)
A 4 class model was selected.
reasons here. Have inserted a place holder of the 5 class model as this is looking best so far. Replace this image with the figure for the best model when that has been identified
best fit model for trajectories of symptoms of depression over one year of the pandemic
We interpret each class as follows:
Class 1 (green line):
Class 2 (green line):
Class 3 (green line):
Class 4 (green line):
Class 5 (green line):
Anxiety (GAD-7)
A 4 class model was selected.
best fit model for trajectories of symptoms of anxiety over one year of the pandemic
We interpret each class as follows:
Class 1 (green line):
Class 2 (green line):
Class 3 (green line):
Class 4 (green line):
Class 5 (green line):
Anhedonia (MASQ)
A x class model was selected.
reasons here. Have inserted a place holder of the 5 class model as this is looking best so far. Replace this image with the figure for the best model when that has been identified
best fit model for trajectories of anhedonia over one year of the pandemic
We interpret each class as follows:
Class 1 (green line):
Class 2 (green line):
Class 3 (green line):
Class 4 (green line):
Class 5 (green line):
Description of selected predictors
The predictors we include which were significant predictors of change in symptoms of anxiety and/or depression in our sample from before the pandemic to the first measurement pint during the pandemic are:
- Gender (being female predicted worsening of symtoms)
- Age (being younger; i.e. <35, predicted worsening of symptoms)
- Employment status pre pandemic (being unemployed predicted worsening of symptoms)
- Pre-existing mental health diagnoses (a range of mental health diagnoses predicted worsening of symptoms)
We include a few additional measures which were not significant predictors in our first paper, but which we feel are important to consider here.
- ethnicity (described in the pre-registration; see note below for more detail)
- key worker status. This was a predictor in our first paper.although we did not plan to include it in the pre-registration, given the piecewise nature of our best fitting latent growth curve based on national pandemic related events, we consider this to be a potential important explanatory variable for differing responses to lockdowns and restrictions easing.
- employment change during early pandemic (increased employment or became employed, decreased employment or became unemployed, furloughed, no change). Whilst we did not use this variable in our earlier paper, given the role for employment n change from pre-pandemic to early pandemic, it is reasonable to infer that any change in employment status during the pandemic may be an important predictor of symptom trajectories during the pandemic.
A note on how we will handle ethnicity as a predictor.
We will use the ethnicity variable in two different ways in order to ensure we are making the best possible use of the data we have available for this important predictor. First, we will use each self reported ethnicity category as an individual predictor. Although each individual group may be too small to be well powered to detect small effects, we feel it is important to recognise that experiences of different minoritised ethnic communities are not homogeneous, and these different experiences may have an important impact on the overall trajectory of mental healthy symptoms during the pandemic.By collapsing across different minoritsed ethnic community groups, we may miss important differences in potential risk factors and outcomes associated with any one specific ethnicity or background experience.
We will then perform a second analysis, where we group all minoritised ethnic communities together, comparing them to the visible majority white ethnicity group. Whist this grouping does not respect the different experiences that different communities may experience, it does consider the impact of being a “visible minority” and the potential risk for discrimination that this may bring. This analysis will be better powered to detect smaller effects.
We hope that by performing this analysis in both ways, we will be able to fully consider the possible impact of experiences of all ethnic communities on anxiety, depression and anhedonia trajectories throughout the pandemic.
SOME EDITS NEEDED WHEN FINAL CLASS DECIDED
Anywhere that the code is dependent on the number of classes in dataset
Analyses
set up
read in all functions from function library
read in class probability and predictor data
The format of the data output saved from Cprobabilities in mplus is:
first section = that persons raw data for each time point (as many columns as there are time points, so GAD and PHQ = 17, MASQ = 15)
Second section = Individuals Intercept, Lin1 slope, lin2 slope lin 3 slope, lin4 slope (always 5 columns) Third section = Class intercept, ine1 slope, lin2 slope, lin3 slope, lin4 slope (always 5 columns) Fourth section = class posterior probability estimate for each estimated class (as many columns as there are classes)
Second last column: most likely class membership Last column: ID
read in the predictor data
##Data preparation ### name and select the class and ID variables for multinomial regression use our custom function rename_class_data
merge predictors with mplus class data
Left join using MPlus hashed ID so we only retain those included in the MPlus model stage.
This doesn’t change the numbers (reassuring)
drop people without data on any predictor
this represents people who were not missing on all variables as per MPLus output file, and our true analytic sample.
Descriptives of the relationship between class and predictors
PHQ
| Characteristic | age | Total | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 16-18 | 19-25 | 26-35 | 36-45 | 46-55 | 56-65 | 66-70 | 71-75 | 76+ | Unknown | ||
| MostLikelyClass | |||||||||||
| 1 | 201 (7.2%) | 517 (19%) | 585 (21%) | 485 (17%) | 625 (22%) | 293 (11%) | 53 (1.9%) | 18 (0.6%) | 5 (0.2%) | 0 (0%) | 2,782 (100%) |
| 2 | 92 (0.4%) | 844 (4.1%) | 2,389 (12%) | 2,772 (14%) | 4,250 (21%) | 5,333 (26%) | 2,483 (12%) | 1,728 (8.4%) | 561 (2.7%) | 7 (<0.1%) | 20,459 (100%) |
| 3 | 125 (2.4%) | 632 (12%) | 1,225 (24%) | 999 (19%) | 1,093 (21%) | 780 (15%) | 179 (3.5%) | 87 (1.7%) | 38 (0.7%) | 0 (0%) | 5,158 (100%) |
| 4 | 148 (4.6%) | 535 (17%) | 832 (26%) | 576 (18%) | 598 (19%) | 403 (13%) | 77 (2.4%) | 37 (1.1%) | 12 (0.4%) | 1 (<0.1%) | 3,219 (100%) |
| Total | 566 (1.8%) | 2,528 (8.0%) | 5,031 (16%) | 4,832 (15%) | 6,566 (21%) | 6,809 (22%) | 2,792 (8.8%) | 1,870 (5.9%) | 616 (1.9%) | 8 (<0.1%) | 31,618 (100%) |
| Characteristic | gender | Total | |||
|---|---|---|---|---|---|
| Male | Female | Non-binary/Prefer to self define | Unknown | ||
| MostLikelyClass | |||||
| 1 | 511 (18%) | 2,161 (78%) | 84 (3.0%) | 26 (0.9%) | 2,782 (100%) |
| 2 | 7,607 (37%) | 12,735 (62%) | 87 (0.4%) | 30 (0.1%) | 20,459 (100%) |
| 3 | 1,020 (20%) | 4,001 (78%) | 110 (2.1%) | 27 (0.5%) | 5,158 (100%) |
| 4 | 510 (16%) | 2,619 (81%) | 76 (2.4%) | 14 (0.4%) | 3,219 (100%) |
| Total | 9,648 (31%) | 21,516 (68%) | 357 (1.1%) | 97 (0.3%) | 31,618 (100%) |
| Characteristic | ethnicity | Total | ||||||
|---|---|---|---|---|---|---|---|---|
| White | Mixed or multiple ethnicity | Asian | Black | Arab | Other | Unknown | ||
| MostLikelyClass | ||||||||
| 1 | 2,584 (93%) | 86 (3.1%) | 30 (1.1%) | 9 (0.3%) | 1 (<0.1%) | 24 (0.9%) | 48 (1.7%) | 2,782 (100%) |
| 2 | 17,080 (83%) | 224 (1.1%) | 173 (0.8%) | 55 (0.3%) | 3 (<0.1%) | 49 (0.2%) | 2,875 (14%) | 20,459 (100%) |
| 3 | 4,700 (91%) | 112 (2.2%) | 51 (1.0%) | 17 (0.3%) | 1 (<0.1%) | 32 (0.6%) | 245 (4.7%) | 5,158 (100%) |
| 4 | 2,956 (92%) | 66 (2.1%) | 34 (1.1%) | 9 (0.3%) | 2 (<0.1%) | 31 (1.0%) | 121 (3.8%) | 3,219 (100%) |
| Total | 27,320 (86%) | 488 (1.5%) | 288 (0.9%) | 90 (0.3%) | 7 (<0.1%) | 136 (0.4%) | 3,289 (10%) | 31,618 (100%) |
| Characteristic | ethnicity_clustered | Total | ||
|---|---|---|---|---|
| White | Minoritised ethnic community | Unknown | ||
| MostLikelyClass | ||||
| 1 | 2,584 (93%) | 150 (5.4%) | 48 (1.7%) | 2,782 (100%) |
| 2 | 17,080 (83%) | 504 (2.5%) | 2,875 (14%) | 20,459 (100%) |
| 3 | 4,700 (91%) | 213 (4.1%) | 245 (4.7%) | 5,158 (100%) |
| 4 | 2,956 (92%) | 142 (4.4%) | 121 (3.8%) | 3,219 (100%) |
| Total | 27,320 (86%) | 1,009 (3.2%) | 3,289 (10%) | 31,618 (100%) |
| Characteristic | employment_prior_covid | Total | ||||
|---|---|---|---|---|---|---|
| Employed | Retired | Student | Unemployed | Unknown | ||
| MostLikelyClass | ||||||
| 1 | 1,275 (46%) | 196 (7.0%) | 226 (8.1%) | 463 (17%) | 622 (22%) | 2,782 (100%) |
| 2 | 12,730 (62%) | 6,127 (30%) | 313 (1.5%) | 465 (2.3%) | 824 (4.0%) | 20,459 (100%) |
| 3 | 3,311 (64%) | 533 (10%) | 237 (4.6%) | 399 (7.7%) | 678 (13%) | 5,158 (100%) |
| 4 | 2,039 (63%) | 242 (7.5%) | 218 (6.8%) | 278 (8.6%) | 442 (14%) | 3,219 (100%) |
| Total | 19,355 (61%) | 7,098 (22%) | 994 (3.1%) | 1,605 (5.1%) | 2,566 (8.1%) | 31,618 (100%) |
| Characteristic | employment_change | Total | ||||
|---|---|---|---|---|---|---|
| Decreased employment | Employment not changed | Furloughed | Increased employment | Unknown | ||
| MostLikelyClass | ||||||
| 1 | 283 (10%) | 1,228 (44%) | 251 (9.0%) | 152 (5.5%) | 868 (31%) | 2,782 (100%) |
| 2 | 1,642 (8.0%) | 8,566 (42%) | 1,875 (9.2%) | 571 (2.8%) | 7,805 (38%) | 20,459 (100%) |
| 3 | 522 (10%) | 2,332 (45%) | 560 (11%) | 254 (4.9%) | 1,490 (29%) | 5,158 (100%) |
| 4 | 383 (12%) | 1,371 (43%) | 363 (11%) | 142 (4.4%) | 960 (30%) | 3,219 (100%) |
| Total | 2,830 (9.0%) | 13,497 (43%) | 3,049 (9.6%) | 1,119 (3.5%) | 11,123 (35%) | 31,618 (100%) |
| Characteristic | key_worker | Total | ||
|---|---|---|---|---|
| Government defined key worker | Not a key worker | Unknown | ||
| MostLikelyClass | ||||
| 1 | 787 (28%) | 1,706 (61%) | 289 (10%) | 2,782 (100%) |
| 2 | 7,410 (36%) | 12,667 (62%) | 382 (1.9%) | 20,459 (100%) |
| 3 | 2,055 (40%) | 2,776 (54%) | 327 (6.3%) | 5,158 (100%) |
| 4 | 1,236 (38%) | 1,720 (53%) | 263 (8.2%) | 3,219 (100%) |
| Total | 11,488 (36%) | 18,869 (60%) | 1,261 (4.0%) | 31,618 (100%) |
| Characteristic | anxiety_disorders | Total | ||
|---|---|---|---|---|
| Anxiety disorder | No anxiety disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 2,332 (84%) | 366 (13%) | 84 (3.0%) | 2,782 (100%) |
| 2 | 6,018 (29%) | 13,873 (68%) | 568 (2.8%) | 20,459 (100%) |
| 3 | 3,632 (70%) | 1,305 (25%) | 221 (4.3%) | 5,158 (100%) |
| 4 | 2,317 (72%) | 738 (23%) | 164 (5.1%) | 3,219 (100%) |
| Total | 14,299 (45%) | 16,282 (51%) | 1,037 (3.3%) | 31,618 (100%) |
| Characteristic | depressive_disorders | Total | ||
|---|---|---|---|---|
| Depressive disorder | No depressive disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 2,549 (92%) | 227 (8.2%) | 6 (0.2%) | 2,782 (100%) |
| 2 | 7,376 (36%) | 13,014 (64%) | 69 (0.3%) | 20,459 (100%) |
| 3 | 4,181 (81%) | 968 (19%) | 9 (0.2%) | 5,158 (100%) |
| 4 | 2,689 (84%) | 526 (16%) | 4 (0.1%) | 3,219 (100%) |
| Total | 16,795 (53%) | 14,735 (47%) | 88 (0.3%) | 31,618 (100%) |
| Characteristic | eating_disorders | Total | ||
|---|---|---|---|---|
| Eating disorder | No eating disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 678 (24%) | 2,095 (75%) | 9 (0.3%) | 2,782 (100%) |
| 2 | 705 (3.4%) | 19,742 (96%) | 12 (<0.1%) | 20,459 (100%) |
| 3 | 686 (13%) | 4,458 (86%) | 14 (0.3%) | 5,158 (100%) |
| 4 | 486 (15%) | 2,730 (85%) | 3 (<0.1%) | 3,219 (100%) |
| Total | 2,555 (8.1%) | 29,025 (92%) | 38 (0.1%) | 31,618 (100%) |
| Characteristic | obsessive_compulsive_disorders | Total | ||
|---|---|---|---|---|
| No obsessive compulsive disorder | Obsessive compulsive disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 2,148 (77%) | 632 (23%) | 2 (<0.1%) | 2,782 (100%) |
| 2 | 19,797 (97%) | 606 (3.0%) | 56 (0.3%) | 20,459 (100%) |
| 3 | 4,536 (88%) | 618 (12%) | 4 (<0.1%) | 5,158 (100%) |
| 4 | 2,778 (86%) | 437 (14%) | 4 (0.1%) | 3,219 (100%) |
| Total | 29,259 (93%) | 2,293 (7.3%) | 66 (0.2%) | 31,618 (100%) |
| Characteristic | psychotic_disorders | Total | ||
|---|---|---|---|---|
| No psychotic disorder | Psychotic disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 2,618 (94%) | 153 (5.5%) | 11 (0.4%) | 2,782 (100%) |
| 2 | 20,215 (99%) | 173 (0.8%) | 71 (0.3%) | 20,459 (100%) |
| 3 | 5,002 (97%) | 133 (2.6%) | 23 (0.4%) | 5,158 (100%) |
| 4 | 3,144 (98%) | 67 (2.1%) | 8 (0.2%) | 3,219 (100%) |
| Total | 30,979 (98%) | 526 (1.7%) | 113 (0.4%) | 31,618 (100%) |
| Characteristic | personality_disorders | Total | ||
|---|---|---|---|---|
| No personality disorder | Personality disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 2,275 (82%) | 496 (18%) | 11 (0.4%) | 2,782 (100%) |
| 2 | 20,172 (99%) | 216 (1.1%) | 71 (0.3%) | 20,459 (100%) |
| 3 | 4,783 (93%) | 352 (6.8%) | 23 (0.4%) | 5,158 (100%) |
| 4 | 2,956 (92%) | 255 (7.9%) | 8 (0.2%) | 3,219 (100%) |
| Total | 30,186 (95%) | 1,319 (4.2%) | 113 (0.4%) | 31,618 (100%) |
| Characteristic | bipolar_disorders | Total | ||
|---|---|---|---|---|
| Bipolar disorder | No bipolar disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 275 (9.9%) | 2,505 (90%) | 2 (<0.1%) | 2,782 (100%) |
| 2 | 330 (1.6%) | 20,073 (98%) | 56 (0.3%) | 20,459 (100%) |
| 3 | 316 (6.1%) | 4,838 (94%) | 4 (<0.1%) | 5,158 (100%) |
| 4 | 188 (5.8%) | 3,027 (94%) | 4 (0.1%) | 3,219 (100%) |
| Total | 1,109 (3.5%) | 30,443 (96%) | 66 (0.2%) | 31,618 (100%) |
GAD
| Characteristic | age | Total | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 16-18 | 19-25 | 26-35 | 36-45 | 46-55 | 56-65 | 66-70 | 71-75 | 76+ | Unknown | ||
| MostLikelyClass | |||||||||||
| 1 | 190 (0.8%) | 1,111 (4.9%) | 2,935 (13%) | 3,138 (14%) | 4,722 (21%) | 5,696 (25%) | 2,558 (11%) | 1,759 (7.8%) | 574 (2.5%) | 6 (<0.1%) | 22,689 (100%) |
| 2 | 221 (5.7%) | 742 (19%) | 936 (24%) | 694 (18%) | 783 (20%) | 388 (10%) | 67 (1.7%) | 24 (0.6%) | 10 (0.3%) | 2 (<0.1%) | 3,867 (100%) |
| 3 | 99 (4.3%) | 356 (15%) | 590 (26%) | 435 (19%) | 426 (18%) | 297 (13%) | 63 (2.7%) | 28 (1.2%) | 13 (0.6%) | 0 (0%) | 2,307 (100%) |
| 4 | 43 (2.0%) | 245 (11%) | 464 (21%) | 459 (21%) | 495 (23%) | 350 (16%) | 84 (3.8%) | 42 (1.9%) | 16 (0.7%) | 0 (0%) | 2,198 (100%) |
| Total | 553 (1.8%) | 2,454 (7.9%) | 4,925 (16%) | 4,726 (15%) | 6,426 (21%) | 6,731 (22%) | 2,772 (8.9%) | 1,853 (6.0%) | 613 (2.0%) | 8 (<0.1%) | 31,061 (100%) |
| Characteristic | gender | Total | |||
|---|---|---|---|---|---|
| Male | Female | Non-binary/Prefer to self define | Unknown | ||
| MostLikelyClass | |||||
| 1 | 8,111 (36%) | 14,365 (63%) | 162 (0.7%) | 51 (0.2%) | 22,689 (100%) |
| 2 | 592 (15%) | 3,132 (81%) | 116 (3.0%) | 27 (0.7%) | 3,867 (100%) |
| 3 | 347 (15%) | 1,917 (83%) | 37 (1.6%) | 6 (0.3%) | 2,307 (100%) |
| 4 | 407 (19%) | 1,740 (79%) | 40 (1.8%) | 11 (0.5%) | 2,198 (100%) |
| Total | 9,457 (30%) | 21,154 (68%) | 355 (1.1%) | 95 (0.3%) | 31,061 (100%) |
| Characteristic | ethnicity | Total | ||||||
|---|---|---|---|---|---|---|---|---|
| White | Mixed or multiple ethnicity | Asian | Black | Arab | Other | Unknown | ||
| MostLikelyClass | ||||||||
| 1 | 19,204 (85%) | 285 (1.3%) | 193 (0.9%) | 59 (0.3%) | 4 (<0.1%) | 64 (0.3%) | 2,880 (13%) | 22,689 (100%) |
| 2 | 3,602 (93%) | 96 (2.5%) | 38 (1.0%) | 9 (0.2%) | 1 (<0.1%) | 34 (0.9%) | 87 (2.2%) | 3,867 (100%) |
| 3 | 2,075 (90%) | 54 (2.3%) | 25 (1.1%) | 8 (0.3%) | 2 (<0.1%) | 27 (1.2%) | 116 (5.0%) | 2,307 (100%) |
| 4 | 2,001 (91%) | 43 (2.0%) | 25 (1.1%) | 14 (0.6%) | 0 (0%) | 11 (0.5%) | 104 (4.7%) | 2,198 (100%) |
| Total | 26,882 (87%) | 478 (1.5%) | 281 (0.9%) | 90 (0.3%) | 7 (<0.1%) | 136 (0.4%) | 3,187 (10%) | 31,061 (100%) |
| Characteristic | ethnicity_clustered | Total | ||
|---|---|---|---|---|
| White | Minoritised ethnic community | Unknown | ||
| MostLikelyClass | ||||
| 1 | 19,204 (85%) | 605 (2.7%) | 2,880 (13%) | 22,689 (100%) |
| 2 | 3,602 (93%) | 178 (4.6%) | 87 (2.2%) | 3,867 (100%) |
| 3 | 2,075 (90%) | 116 (5.0%) | 116 (5.0%) | 2,307 (100%) |
| 4 | 2,001 (91%) | 93 (4.2%) | 104 (4.7%) | 2,198 (100%) |
| Total | 26,882 (87%) | 992 (3.2%) | 3,187 (10%) | 31,061 (100%) |
| Characteristic | employment_prior_covid | Total | ||||
|---|---|---|---|---|---|---|
| Employed | Retired | Student | Unemployed | Unknown | ||
| MostLikelyClass | ||||||
| 1 | 13,986 (62%) | 6,349 (28%) | 436 (1.9%) | 758 (3.3%) | 1,160 (5.1%) | 22,689 (100%) |
| 2 | 2,075 (54%) | 250 (6.5%) | 298 (7.7%) | 531 (14%) | 713 (18%) | 3,867 (100%) |
| 3 | 1,510 (65%) | 185 (8.0%) | 145 (6.3%) | 152 (6.6%) | 315 (14%) | 2,307 (100%) |
| 4 | 1,474 (67%) | 247 (11%) | 107 (4.9%) | 147 (6.7%) | 223 (10%) | 2,198 (100%) |
| Total | 19,045 (61%) | 7,031 (23%) | 986 (3.2%) | 1,588 (5.1%) | 2,411 (7.8%) | 31,061 (100%) |
| Characteristic | employment_change | Total | ||||
|---|---|---|---|---|---|---|
| Decreased employment | Employment not changed | Furloughed | Increased employment | Unknown | ||
| MostLikelyClass | ||||||
| 1 | 1,886 (8.3%) | 9,545 (42%) | 2,100 (9.3%) | 667 (2.9%) | 8,491 (37%) | 22,689 (100%) |
| 2 | 419 (11%) | 1,671 (43%) | 383 (9.9%) | 230 (5.9%) | 1,164 (30%) | 3,867 (100%) |
| 3 | 264 (11%) | 1,027 (45%) | 258 (11%) | 107 (4.6%) | 651 (28%) | 2,307 (100%) |
| 4 | 230 (10%) | 1,043 (47%) | 254 (12%) | 94 (4.3%) | 577 (26%) | 2,198 (100%) |
| Total | 2,799 (9.0%) | 13,286 (43%) | 2,995 (9.6%) | 1,098 (3.5%) | 10,883 (35%) | 31,061 (100%) |
| Characteristic | key_worker | Total | ||
|---|---|---|---|---|
| Government defined key worker | Not a key worker | Unknown | ||
| MostLikelyClass | ||||
| 1 | 8,184 (36%) | 13,965 (62%) | 540 (2.4%) | 22,689 (100%) |
| 2 | 1,271 (33%) | 2,245 (58%) | 351 (9.1%) | 3,867 (100%) |
| 3 | 925 (40%) | 1,217 (53%) | 165 (7.2%) | 2,307 (100%) |
| 4 | 930 (42%) | 1,198 (55%) | 70 (3.2%) | 2,198 (100%) |
| Total | 11,310 (36%) | 18,625 (60%) | 1,126 (3.6%) | 31,061 (100%) |
| Characteristic | anxiety_disorders | Total | ||
|---|---|---|---|---|
| Anxiety disorder | No anxiety disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 7,435 (33%) | 14,491 (64%) | 763 (3.4%) | 22,689 (100%) |
| 2 | 3,312 (86%) | 474 (12%) | 81 (2.1%) | 3,867 (100%) |
| 3 | 1,694 (73%) | 511 (22%) | 102 (4.4%) | 2,307 (100%) |
| 4 | 1,635 (74%) | 483 (22%) | 80 (3.6%) | 2,198 (100%) |
| Total | 14,076 (45%) | 15,959 (51%) | 1,026 (3.3%) | 31,061 (100%) |
| Characteristic | depressive_disorders | Total | ||
|---|---|---|---|---|
| Depressive disorder | No depressive disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 9,595 (42%) | 13,028 (57%) | 66 (0.3%) | 22,689 (100%) |
| 2 | 3,360 (87%) | 499 (13%) | 8 (0.2%) | 3,867 (100%) |
| 3 | 1,841 (80%) | 463 (20%) | 3 (0.1%) | 2,307 (100%) |
| 4 | 1,764 (80%) | 433 (20%) | 1 (<0.1%) | 2,198 (100%) |
| Total | 16,560 (53%) | 14,423 (46%) | 78 (0.3%) | 31,061 (100%) |
| Characteristic | eating_disorders | Total | ||
|---|---|---|---|---|
| Eating disorder | No eating disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 1,024 (4.5%) | 21,644 (95%) | 21 (<0.1%) | 22,689 (100%) |
| 2 | 858 (22%) | 3,002 (78%) | 7 (0.2%) | 3,867 (100%) |
| 3 | 350 (15%) | 1,954 (85%) | 3 (0.1%) | 2,307 (100%) |
| 4 | 281 (13%) | 1,913 (87%) | 4 (0.2%) | 2,198 (100%) |
| Total | 2,513 (8.1%) | 28,513 (92%) | 35 (0.1%) | 31,061 (100%) |
| Characteristic | obsessive_compulsive_disorders | Total | ||
|---|---|---|---|---|
| No obsessive compulsive disorder | Obsessive compulsive disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 21,781 (96%) | 855 (3.8%) | 53 (0.2%) | 22,689 (100%) |
| 2 | 3,049 (79%) | 814 (21%) | 4 (0.1%) | 3,867 (100%) |
| 3 | 1,991 (86%) | 314 (14%) | 2 (<0.1%) | 2,307 (100%) |
| 4 | 1,922 (87%) | 276 (13%) | 0 (0%) | 2,198 (100%) |
| Total | 28,743 (93%) | 2,259 (7.3%) | 59 (0.2%) | 31,061 (100%) |
| Characteristic | psychotic_disorders | Total | ||
|---|---|---|---|---|
| No psychotic disorder | Psychotic disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 22,383 (99%) | 234 (1.0%) | 72 (0.3%) | 22,689 (100%) |
| 2 | 3,692 (95%) | 163 (4.2%) | 12 (0.3%) | 3,867 (100%) |
| 3 | 2,248 (97%) | 51 (2.2%) | 8 (0.3%) | 2,307 (100%) |
| 4 | 2,120 (96%) | 70 (3.2%) | 8 (0.4%) | 2,198 (100%) |
| Total | 30,443 (98%) | 518 (1.7%) | 100 (0.3%) | 31,061 (100%) |
| Characteristic | personality_disorders | Total | ||
|---|---|---|---|---|
| No personality disorder | Personality disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 22,203 (98%) | 414 (1.8%) | 72 (0.3%) | 22,689 (100%) |
| 2 | 3,302 (85%) | 553 (14%) | 12 (0.3%) | 3,867 (100%) |
| 3 | 2,138 (93%) | 161 (7.0%) | 8 (0.3%) | 2,307 (100%) |
| 4 | 2,024 (92%) | 166 (7.6%) | 8 (0.4%) | 2,198 (100%) |
| Total | 29,667 (96%) | 1,294 (4.2%) | 100 (0.3%) | 31,061 (100%) |
| Characteristic | bipolar_disorders | Total | ||
|---|---|---|---|---|
| Bipolar disorder | No bipolar disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 513 (2.3%) | 22,123 (98%) | 53 (0.2%) | 22,689 (100%) |
| 2 | 321 (8.3%) | 3,542 (92%) | 4 (0.1%) | 3,867 (100%) |
| 3 | 123 (5.3%) | 2,182 (95%) | 2 (<0.1%) | 2,307 (100%) |
| 4 | 135 (6.1%) | 2,063 (94%) | 0 (0%) | 2,198 (100%) |
| Total | 1,092 (3.5%) | 29,910 (96%) | 59 (0.2%) | 31,061 (100%) |
MASQ
| Characteristic | age | Total | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 16-18 | 19-25 | 26-35 | 36-45 | 46-55 | 56-65 | 66-70 | 71-75 | 76+ | Unknown | ||
| MostLikelyClass | |||||||||||
| 1 | 8 (1.7%) | 32 (6.9%) | 89 (19%) | 91 (20%) | 100 (22%) | 90 (19%) | 28 (6.0%) | 18 (3.9%) | 6 (1.3%) | 1 (0.2%) | 463 (100%) |
| 2 | 45 (0.6%) | 333 (4.7%) | 835 (12%) | 859 (12%) | 1,319 (18%) | 1,861 (26%) | 988 (14%) | 693 (9.7%) | 214 (3.0%) | 2 (<0.1%) | 7,149 (100%) |
| 3 | 289 (1.7%) | 1,187 (7.0%) | 2,604 (15%) | 2,772 (16%) | 3,806 (22%) | 3,788 (22%) | 1,384 (8.1%) | 907 (5.3%) | 297 (1.7%) | 3 (<0.1%) | 17,037 (100%) |
| 4 | 7 (1.1%) | 55 (8.8%) | 101 (16%) | 95 (15%) | 138 (22%) | 137 (22%) | 61 (9.8%) | 22 (3.5%) | 8 (1.3%) | 0 (0%) | 624 (100%) |
| 5 | 12 (2.4%) | 43 (8.4%) | 96 (19%) | 72 (14%) | 109 (21%) | 109 (21%) | 38 (7.5%) | 21 (4.1%) | 9 (1.8%) | 0 (0%) | 509 (100%) |
| Total | 361 (1.4%) | 1,650 (6.4%) | 3,725 (14%) | 3,889 (15%) | 5,472 (21%) | 5,985 (23%) | 2,499 (9.7%) | 1,661 (6.4%) | 534 (2.1%) | 6 (<0.1%) | 25,782 (100%) |
| Characteristic | gender | Total | |||
|---|---|---|---|---|---|
| Male | Female | Non-binary/Prefer to self define | Unknown | ||
| MostLikelyClass | |||||
| 1 | 127 (27%) | 329 (71%) | 6 (1.3%) | 1 (0.2%) | 463 (100%) |
| 2 | 3,128 (44%) | 3,966 (55%) | 45 (0.6%) | 10 (0.1%) | 7,149 (100%) |
| 3 | 4,481 (26%) | 12,278 (72%) | 220 (1.3%) | 58 (0.3%) | 17,037 (100%) |
| 4 | 120 (19%) | 499 (80%) | 2 (0.3%) | 3 (0.5%) | 624 (100%) |
| 5 | 99 (19%) | 400 (79%) | 7 (1.4%) | 3 (0.6%) | 509 (100%) |
| Total | 7,955 (31%) | 17,472 (68%) | 280 (1.1%) | 75 (0.3%) | 25,782 (100%) |
| Characteristic | ethnicity | Total | ||||||
|---|---|---|---|---|---|---|---|---|
| White | Mixed or multiple ethnicity | Asian | Black | Arab | Other | Unknown | ||
| MostLikelyClass | ||||||||
| 1 | 403 (87%) | 9 (1.9%) | 4 (0.9%) | 1 (0.2%) | 0 (0%) | 3 (0.6%) | 43 (9.3%) | 463 (100%) |
| 2 | 5,892 (82%) | 82 (1.1%) | 65 (0.9%) | 22 (0.3%) | 1 (<0.1%) | 15 (0.2%) | 1,072 (15%) | 7,149 (100%) |
| 3 | 15,152 (89%) | 254 (1.5%) | 124 (0.7%) | 48 (0.3%) | 5 (<0.1%) | 87 (0.5%) | 1,367 (8.0%) | 17,037 (100%) |
| 4 | 524 (84%) | 11 (1.8%) | 9 (1.4%) | 2 (0.3%) | 0 (0%) | 1 (0.2%) | 77 (12%) | 624 (100%) |
| 5 | 451 (89%) | 10 (2.0%) | 7 (1.4%) | 0 (0%) | 0 (0%) | 5 (1.0%) | 36 (7.1%) | 509 (100%) |
| Total | 22,422 (87%) | 366 (1.4%) | 209 (0.8%) | 73 (0.3%) | 6 (<0.1%) | 111 (0.4%) | 2,595 (10%) | 25,782 (100%) |
| Characteristic | ethnicity_clustered | Total | ||
|---|---|---|---|---|
| White | Minoritised ethnic community | Unknown | ||
| MostLikelyClass | ||||
| 1 | 403 (87%) | 17 (3.7%) | 43 (9.3%) | 463 (100%) |
| 2 | 5,892 (82%) | 185 (2.6%) | 1,072 (15%) | 7,149 (100%) |
| 3 | 15,152 (89%) | 518 (3.0%) | 1,367 (8.0%) | 17,037 (100%) |
| 4 | 524 (84%) | 23 (3.7%) | 77 (12%) | 624 (100%) |
| 5 | 451 (89%) | 22 (4.3%) | 36 (7.1%) | 509 (100%) |
| Total | 22,422 (87%) | 765 (3.0%) | 2,595 (10%) | 25,782 (100%) |
| Characteristic | employment_prior_covid | Total | ||||
|---|---|---|---|---|---|---|
| Employed | Retired | Student | Unemployed | Unknown | ||
| MostLikelyClass | ||||||
| 1 | 331 (71%) | 75 (16%) | 16 (3.5%) | 23 (5.0%) | 18 (3.9%) | 463 (100%) |
| 2 | 4,386 (61%) | 2,335 (33%) | 110 (1.5%) | 139 (1.9%) | 179 (2.5%) | 7,149 (100%) |
| 3 | 10,544 (62%) | 3,694 (22%) | 582 (3.4%) | 1,109 (6.5%) | 1,108 (6.5%) | 17,037 (100%) |
| 4 | 417 (67%) | 141 (23%) | 28 (4.5%) | 14 (2.2%) | 24 (3.8%) | 624 (100%) |
| 5 | 351 (69%) | 103 (20%) | 17 (3.3%) | 25 (4.9%) | 13 (2.6%) | 509 (100%) |
| Total | 16,029 (62%) | 6,348 (25%) | 753 (2.9%) | 1,310 (5.1%) | 1,342 (5.2%) | 25,782 (100%) |
| Characteristic | employment_change | Total | ||||
|---|---|---|---|---|---|---|
| Decreased employment | Employment not changed | Furloughed | Increased employment | Unknown | ||
| MostLikelyClass | ||||||
| 1 | 54 (12%) | 225 (49%) | 44 (9.5%) | 14 (3.0%) | 126 (27%) | 463 (100%) |
| 2 | 554 (7.7%) | 2,915 (41%) | 668 (9.3%) | 201 (2.8%) | 2,811 (39%) | 7,149 (100%) |
| 3 | 1,637 (9.6%) | 7,659 (45%) | 1,601 (9.4%) | 644 (3.8%) | 5,496 (32%) | 17,037 (100%) |
| 4 | 68 (11%) | 248 (40%) | 71 (11%) | 23 (3.7%) | 214 (34%) | 624 (100%) |
| 5 | 69 (14%) | 199 (39%) | 63 (12%) | 23 (4.5%) | 155 (30%) | 509 (100%) |
| Total | 2,382 (9.2%) | 11,246 (44%) | 2,447 (9.5%) | 905 (3.5%) | 8,802 (34%) | 25,782 (100%) |
| Characteristic | key_worker | Total | ||
|---|---|---|---|---|
| Government defined key worker | Not a key worker | Unknown | ||
| MostLikelyClass | ||||
| 1 | 186 (40%) | 277 (60%) | 0 (0%) | 463 (100%) |
| 2 | 2,569 (36%) | 4,536 (63%) | 44 (0.6%) | 7,149 (100%) |
| 3 | 6,295 (37%) | 10,551 (62%) | 191 (1.1%) | 17,037 (100%) |
| 4 | 253 (41%) | 366 (59%) | 5 (0.8%) | 624 (100%) |
| 5 | 209 (41%) | 299 (59%) | 1 (0.2%) | 509 (100%) |
| Total | 9,512 (37%) | 16,029 (62%) | 241 (0.9%) | 25,782 (100%) |
| Characteristic | anxiety_disorders | Total | ||
|---|---|---|---|---|
| Anxiety disorder | No anxiety disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 243 (52%) | 192 (41%) | 28 (6.0%) | 463 (100%) |
| 2 | 1,716 (24%) | 5,290 (74%) | 143 (2.0%) | 7,149 (100%) |
| 3 | 8,875 (52%) | 7,529 (44%) | 633 (3.7%) | 17,037 (100%) |
| 4 | 259 (42%) | 342 (55%) | 23 (3.7%) | 624 (100%) |
| 5 | 253 (50%) | 232 (46%) | 24 (4.7%) | 509 (100%) |
| Total | 11,346 (44%) | 13,585 (53%) | 851 (3.3%) | 25,782 (100%) |
| Characteristic | depressive_disorders | Total | ||
|---|---|---|---|---|
| Depressive disorder | No depressive disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 305 (66%) | 157 (34%) | 1 (0.2%) | 463 (100%) |
| 2 | 2,002 (28%) | 5,124 (72%) | 23 (0.3%) | 7,149 (100%) |
| 3 | 10,571 (62%) | 6,436 (38%) | 30 (0.2%) | 17,037 (100%) |
| 4 | 303 (49%) | 320 (51%) | 1 (0.2%) | 624 (100%) |
| 5 | 318 (62%) | 190 (37%) | 1 (0.2%) | 509 (100%) |
| Total | 13,499 (52%) | 12,227 (47%) | 56 (0.2%) | 25,782 (100%) |
| Characteristic | eating_disorders | Total | ||
|---|---|---|---|---|
| Eating disorder | No eating disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 42 (9.1%) | 420 (91%) | 1 (0.2%) | 463 (100%) |
| 2 | 222 (3.1%) | 6,922 (97%) | 5 (<0.1%) | 7,149 (100%) |
| 3 | 1,569 (9.2%) | 15,453 (91%) | 15 (<0.1%) | 17,037 (100%) |
| 4 | 46 (7.4%) | 577 (92%) | 1 (0.2%) | 624 (100%) |
| 5 | 52 (10%) | 457 (90%) | 0 (0%) | 509 (100%) |
| Total | 1,931 (7.5%) | 23,829 (92%) | 22 (<0.1%) | 25,782 (100%) |
| Characteristic | obsessive_compulsive_disorders | Total | ||
|---|---|---|---|---|
| No obsessive compulsive disorder | Obsessive compulsive disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 440 (95%) | 22 (4.8%) | 1 (0.2%) | 463 (100%) |
| 2 | 6,914 (97%) | 213 (3.0%) | 22 (0.3%) | 7,149 (100%) |
| 3 | 15,564 (91%) | 1,454 (8.5%) | 19 (0.1%) | 17,037 (100%) |
| 4 | 586 (94%) | 37 (5.9%) | 1 (0.2%) | 624 (100%) |
| 5 | 483 (95%) | 25 (4.9%) | 1 (0.2%) | 509 (100%) |
| Total | 23,987 (93%) | 1,751 (6.8%) | 44 (0.2%) | 25,782 (100%) |
| Characteristic | psychotic_disorders | Total | ||
|---|---|---|---|---|
| No psychotic disorder | Psychotic disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 457 (99%) | 5 (1.1%) | 1 (0.2%) | 463 (100%) |
| 2 | 7,064 (99%) | 60 (0.8%) | 25 (0.3%) | 7,149 (100%) |
| 3 | 16,662 (98%) | 330 (1.9%) | 45 (0.3%) | 17,037 (100%) |
| 4 | 608 (97%) | 12 (1.9%) | 4 (0.6%) | 624 (100%) |
| 5 | 496 (97%) | 13 (2.6%) | 0 (0%) | 509 (100%) |
| Total | 25,287 (98%) | 420 (1.6%) | 75 (0.3%) | 25,782 (100%) |
| Characteristic | personality_disorders | Total | ||
|---|---|---|---|---|
| No personality disorder | Personality disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 439 (95%) | 23 (5.0%) | 1 (0.2%) | 463 (100%) |
| 2 | 7,061 (99%) | 63 (0.9%) | 25 (0.3%) | 7,149 (100%) |
| 3 | 16,136 (95%) | 856 (5.0%) | 45 (0.3%) | 17,037 (100%) |
| 4 | 604 (97%) | 16 (2.6%) | 4 (0.6%) | 624 (100%) |
| 5 | 486 (95%) | 23 (4.5%) | 0 (0%) | 509 (100%) |
| Total | 24,726 (96%) | 981 (3.8%) | 75 (0.3%) | 25,782 (100%) |
| Characteristic | bipolar_disorders | Total | ||
|---|---|---|---|---|
| Bipolar disorder | No bipolar disorder | Unknown | ||
| MostLikelyClass | ||||
| 1 | 28 (6.0%) | 434 (94%) | 1 (0.2%) | 463 (100%) |
| 2 | 91 (1.3%) | 7,036 (98%) | 22 (0.3%) | 7,149 (100%) |
| 3 | 712 (4.2%) | 16,306 (96%) | 19 (0.1%) | 17,037 (100%) |
| 4 | 16 (2.6%) | 607 (97%) | 1 (0.2%) | 624 (100%) |
| 5 | 25 (4.9%) | 483 (95%) | 1 (0.2%) | 509 (100%) |
| Total | 872 (3.4%) | 24,866 (96%) | 44 (0.2%) | 25,782 (100%) |
specify reference category for predictors
make age midpoint the reference for age, no change in employment the reference for employment change, all disorders have the no disorder group as reference, not a keyworker reference for key worker status
# PHQ
phq.AllData$age <- factor(phq.AllData$age,
levels=c("16-18","19-25","26-35","36-45","46-55","56-65","66-70","71-75","76+")) %>%
relevel(phq.AllData$age, ref = "26-35")
phq.AllData$employment_change <- factor(phq.AllData$employment_change,
levels=c("Decreased employment","Employment not changed","Furloughed","Increased employment")) %>%
relevel(phq.AllData$employment_change, ref = "Employment not changed")
phq.AllData$key_worker <- factor(phq.AllData$key_worker,
levels=c("Government defined key worker","Not a key worker")) %>%
relevel(phq.AllData$key_worker, ref = "Not a key worker")
phq.AllData$anxiety_disorders <- factor(phq.AllData$anxiety_disorders,
levels=c("Anxiety disorder","No anxiety disorder")) %>%
relevel(phq.AllData$anxiety_disorders, ref = "No anxiety disorder")
phq.AllData$depressive_disorders <- factor(phq.AllData$depressive_disorders,
levels=c("Depressive disorder","No depressive disorder")) %>%
relevel(phq.AllData$depressive_disorders, ref = "No depressive disorder")
phq.AllData$eating_disorders <- factor(phq.AllData$eating_disorders,
levels=c("Eating disorder","No eating disorder")) %>%
relevel(phq.AllData$eating_disorders, ref = "No eating disorder")
phq.AllData$obsessive_compulsive_disorders <- factor(phq.AllData$obsessive_compulsive_disorders,
levels=c("Obsessive compulsive disorder","No obsessive compulsive disorder")) %>%
relevel(phq.AllData$obsessive_compulsive_disorders, ref = "No obsessive compulsive disorder")
phq.AllData$psychotic_disorders <- factor(phq.AllData$psychotic_disorders,
levels=c("Psychotic disorder","No psychotic disorder")) %>%
relevel(phq.AllData$psychotic_disorders, ref = "No psychotic disorder")
phq.AllData$personality_disorders <- factor(phq.AllData$personality_disorders,
levels=c("Personality disorder","No personality disorder")) %>%
relevel(phq.AllData$personality_disorders, ref = "No personality disorder")
phq.AllData$bipolar_disorders <- factor(phq.AllData$bipolar_disorders,
levels=c("Bipolar disorder","No bipolar disorder")) %>%
relevel(phq.AllData$bipolar_disorders, ref = "No bipolar disorder")
# GAD
gad.AllData$age <- factor(gad.AllData$age,
levels=c("16-18","19-25","26-35","36-45","46-55","56-65","66-70","71-75","76+")) %>%
relevel(gad.AllData$age, ref = "26-35")
gad.AllData$employment_change <- factor(gad.AllData$employment_change,
levels=c("Decreased employment","Employment not changed","Furloughed","Increased employment")) %>%
relevel(gad.AllData$employment_change, ref = "Employment not changed")
gad.AllData$key_worker <- factor(gad.AllData$key_worker,
levels=c("Government defined key worker","Not a key worker")) %>%
relevel(gad.AllData$key_worker, ref = "Not a key worker")
gad.AllData$anxiety_disorders <- factor(gad.AllData$anxiety_disorders,
levels=c("Anxiety disorder","No anxiety disorder")) %>%
relevel(gad.AllData$anxiety_disorders, ref = "No anxiety disorder")
gad.AllData$depressive_disorders <- factor(gad.AllData$depressive_disorders,
levels=c("Depressive disorder","No depressive disorder")) %>%
relevel(gad.AllData$depressive_disorders, ref = "No depressive disorder")
gad.AllData$eating_disorders <- factor(gad.AllData$eating_disorders,
levels=c("Eating disorder","No eating disorder")) %>%
relevel(gad.AllData$eating_disorders, ref = "No eating disorder")
gad.AllData$obsessive_compulsive_disorders <- factor(gad.AllData$obsessive_compulsive_disorders,
levels=c("Obsessive compulsive disorder","No obsessive compulsive disorder")) %>%
relevel(gad.AllData$obsessive_compulsive_disorders, ref = "No obsessive compulsive disorder")
gad.AllData$psychotic_disorders <- factor(gad.AllData$psychotic_disorders,
levels=c("Psychotic disorder","No psychotic disorder")) %>%
relevel(gad.AllData$psychotic_disorders, ref = "No psychotic disorder")
gad.AllData$personality_disorders <- factor(gad.AllData$personality_disorders,
levels=c("Personality disorder","No personality disorder")) %>%
relevel(gad.AllData$personality_disorders, ref = "No personality disorder")
gad.AllData$bipolar_disorders <- factor(gad.AllData$bipolar_disorders,
levels=c("Bipolar disorder","No bipolar disorder")) %>%
relevel(gad.AllData$bipolar_disorders, ref = "No bipolar disorder")
# MASQ
masq.AllData$age <- factor(masq.AllData$age,
levels=c("16-18","19-25","26-35","36-45","46-55","56-65","66-70","71-75","76+")) %>%
relevel(masq.AllData$age, ref = "26-35")
masq.AllData$employment_change <- factor(masq.AllData$employment_change,
levels=c("Decreased employment","Employment not changed","Furloughed","Increased employment")) %>%
relevel(masq.AllData$employment_change, ref = "Employment not changed")
masq.AllData$key_worker <- factor(masq.AllData$key_worker,
levels=c("Government defined key worker","Not a key worker")) %>%
relevel(masq.AllData$key_worker, ref = "Not a key worker")
masq.AllData$anxiety_disorders <- factor(masq.AllData$anxiety_disorders,
levels=c("Anxiety disorder","No anxiety disorder")) %>%
relevel(masq.AllData$anxiety_disorders, ref = "No anxiety disorder")
masq.AllData$depressive_disorders <- factor(masq.AllData$depressive_disorders,
levels=c("Depressive disorder","No depressive disorder")) %>%
relevel(masq.AllData$depressive_disorders, ref = "No depressive disorder")
masq.AllData$eating_disorders <- factor(masq.AllData$eating_disorders,
levels=c("Eating disorder","No eating disorder")) %>%
relevel(masq.AllData$eating_disorders, ref = "No eating disorder")
masq.AllData$obsessive_compulsive_disorders <- factor(masq.AllData$obsessive_compulsive_disorders,
levels=c("Obsessive compulsive disorder","No obsessive compulsive disorder")) %>%
relevel(masq.AllData$obsessive_compulsive_disorders, ref = "No obsessive compulsive disorder")
masq.AllData$psychotic_disorders <- factor(masq.AllData$psychotic_disorders,
levels=c("Psychotic disorder","No psychotic disorder")) %>%
relevel(masq.AllData$psychotic_disorders, ref = "No psychotic disorder")
masq.AllData$personality_disorders <- factor(masq.AllData$personality_disorders,
levels=c("Personality disorder","No personality disorder")) %>%
relevel(masq.AllData$personality_disorders, ref = "No personality disorder")
masq.AllData$bipolar_disorders <- factor(masq.AllData$bipolar_disorders,
levels=c("Bipolar disorder","No bipolar disorder")) %>%
relevel(masq.AllData$bipolar_disorders, ref = "No bipolar disorder")Multinomial regression
formulas
the formulas will be the same for all disorders so create these once here
formula_clustered <- "MostLikelyClass ~ age + gender + ethnicity_clustered + employment_prior_covid + employment_change +key_worker + anxiety_disorders + depressive_disorders + eating_disorders +obsessive_compulsive_disorders +psychotic_disorders + personality_disorders + bipolar_disorders"
formula <- "MostLikelyClass ~ age + gender + ethnicity + employment_prior_covid + employment_change +key_worker + anxiety_disorders + depressive_disorders + eating_disorders +obsessive_compulsive_disorders +psychotic_disorders + personality_disorders + bipolar_disorders"PHQ
prepare and run models
specify reference category for outcome
will use the largest class group as reference
BiggestClass.phq <- phq.AllData %>%
group_by(MostLikelyClass) %>%
summarise(n=n()) %>%
arrange(-n) %>%
slice(1)%>%
select(MostLikelyClass) %>%
as.character(.)
BiggestClass.phq[1] "2"Will need to adapt this to the correct number of classes depending on final model
Perform regression (ethnicity clustered)
# weights: 108 (78 variable)
initial value 22776.816353
iter 10 value 15413.454463
iter 20 value 15018.010656
iter 30 value 14829.289484
iter 40 value 14788.839015
iter 50 value 14780.869242
iter 60 value 14778.570012
iter 70 value 14777.735248
iter 80 value 14777.582866
iter 90 value 14777.495686
final value 14777.495046
convergedPerform regression (ethnicity as is)
# weights: 124 (90 variable)
initial value 22776.816353
iter 10 value 15411.839789
iter 20 value 15013.647665
iter 30 value 14829.869090
iter 40 value 14785.994207
iter 50 value 14778.358654
iter 60 value 14775.657187
iter 70 value 14774.939734
iter 80 value 14774.798621
iter 90 value 14774.630785
iter 100 value 14774.597139
final value 14774.597139
stopped after 100 iterationsView and process results
1: coefficients 2: std errors 3: residual deviances & AIC ##### view summary: Ethnicity clustered
Call:
multinom(formula = formula_clustered, data = phq.AllData)
Coefficients:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 -4.549853 2.1048558 0.7790551 -0.3149913 -0.2498514 -0.9543816 -2.341322
3 -2.547683 0.9017894 0.3711199 -0.2213567 -0.3897026 -0.7070398 -1.121361
4 -3.140364 1.3448804 0.5455621 -0.3948331 -0.6598507 -1.0175320 -1.234575
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 -1.019484 -10.730111 -0.02542089 0.5807275
3 -1.453851 -1.269494 0.16522410 0.5840882
4 -2.661366 -1.382587 0.30206855 0.7226346
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.3708296 1.7704825
3 0.2395683 0.6350815
4 0.2208408 1.2337246
employment_prior_covidStudent employment_prior_covidUnemployed
1 1.8104805 1.960040
3 0.4844898 1.036247
4 0.8510254 1.242283
employment_changeDecreased employment employment_changeFurloughed
1 0.5650230 0.4106201
3 0.3115720 0.2948073
4 0.5195315 0.2970980
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.13575139 0.2522181
3 0.19257965 0.1636542
4 0.02743296 0.1640294
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.9085664 1.948273
3 0.6100999 1.209666
4 0.5927057 1.287837
eating_disordersEating disorder
1 0.9211555
3 0.5823767
4 0.5287970
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.5989446
3 0.2161699
4 0.3800703
psychotic_disordersPsychotic disorder
1 -0.1139183
3 -0.1722263
4 -0.5338968
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 1.0722802 0.6964569
3 0.4741771 0.5781367
4 0.4548947 0.5946864
Std. Errors:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 0.14816774 0.2316826 0.10736351 0.09418929 0.08985454 0.12326540 0.6038183
3 0.08507648 0.2312577 0.08919550 0.06657359 0.06493759 0.07797673 0.2208516
4 0.10673324 0.2304602 0.09590673 0.07867265 0.07868899 0.09853262 0.2723463
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 0.4966426 7.368934e-06 0.08179817 0.2451186
3 0.3554065 6.205729e-01 0.05707060 0.2163430
4 0.7222626 7.493670e-01 0.07267529 0.2381598
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.1488697 0.3823456
3 0.1139642 0.3232712
4 0.1338969 0.3511517
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.4933996 0.10826932
3 0.5260395 0.09334759
4 0.5513469 0.10525114
employment_changeDecreased employment employment_changeFurloughed
1 0.09267171 0.09352993
3 0.06770352 0.06549133
4 0.07816144 0.07892212
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.12813110 0.07569931
3 0.09580487 0.05127901
4 0.11850412 0.06199946
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.09314783 0.11596604
3 0.05791695 0.06135151
4 0.07176675 0.07758567
eating_disordersEating disorder
1 0.09503508
3 0.08237463
4 0.09344419
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.09393152
3 0.08383057
4 0.09228084
psychotic_disordersPsychotic disorder
1 0.1921388
3 0.1726906
4 0.2178653
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.1274057 0.1391339
3 0.1236615 0.1230321
4 0.1371931 0.1400993
Residual Deviance: 29554.99
AIC: 29710.99 view summary: Ethnicity as is
Call:
multinom(formula = formula, data = phq.AllData)
Coefficients:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 -4.553307 2.1039961 0.7796502 -0.3135582 -0.2467409 -0.9515374 -2.336530
3 -2.547118 0.9002638 0.3705796 -0.2201190 -0.3894447 -0.7063024 -1.119318
4 -3.134719 1.3428491 0.5459123 -0.3963309 -0.6621591 -1.0204405 -1.239464
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 -1.017143 -12.224352 -0.02480251 0.5807562
3 -1.453370 -1.270669 0.16369397 0.5807002
4 -2.662583 -1.382245 0.30174999 0.7205345
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
1 0.4629838 0.40858586 0.2066424
3 0.3457871 0.12366733 0.1582022
4 0.1690055 0.08441411 0.3413721
ethnicityArab ethnicityOther employment_prior_covidRetired
1 -5.662052 0.1570573 1.7674622
3 1.029187 0.1020597 0.6327436
4 -7.525484 0.5425931 1.2358961
employment_prior_covidStudent employment_prior_covidUnemployed
1 1.8080856 1.961047
3 0.4822341 1.037672
4 0.8539063 1.240262
employment_changeDecreased employment employment_changeFurloughed
1 0.5664661 0.4110791
3 0.3117479 0.2944403
4 0.5180671 0.2954592
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.13930698 0.2508914
3 0.19402322 0.1642079
4 0.02349589 0.1640721
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.9106050 1.947505
3 0.6097613 1.209428
4 0.5902519 1.286131
eating_disordersEating disorder
1 0.9218347
3 0.5834878
4 0.5270773
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.5979472
3 0.2154201
4 0.3810171
psychotic_disordersPsychotic disorder
1 -0.1152364
3 -0.1729216
4 -0.5311309
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 1.0726504 0.6981503
3 0.4744518 0.5786918
4 0.4551671 0.5916305
Std. Errors:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 0.14831580 0.2316582 0.10738478 0.09423458 0.08991647 0.12331345 0.6036621
3 0.08517708 0.2312299 0.08922588 0.06658966 0.06497050 0.07801082 0.2208580
4 0.10682035 0.2305134 0.09593152 0.07869585 0.07872043 0.09856665 0.2724340
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 0.4967691 1.749907e-06 0.08183394 0.2452735
3 0.3554055 6.209694e-01 0.05709798 0.2163849
4 0.7215908 7.480872e-01 0.07270660 0.2382940
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
1 0.2016425 0.2955720 0.5253943
3 0.1572987 0.2192610 0.3860519
4 0.1909652 0.2666643 0.4347993
ethnicityArab ethnicityOther employment_prior_covidRetired
1 49.416999 0.3884027 0.3823841
3 1.416853 0.3144694 0.3233035
4 85.478333 0.3207641 0.3511165
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.4934754 0.1082854
3 0.5261723 0.0933672
4 0.5513879 0.1053026
employment_changeDecreased employment employment_changeFurloughed
1 0.09268309 0.09356790
3 0.06770987 0.06551691
4 0.07818836 0.07894781
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.12822434 0.07572031
3 0.09582231 0.05129569
4 0.11857777 0.06202039
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.09319813 0.11601099
3 0.05796675 0.06137905
4 0.07182591 0.07762486
eating_disordersEating disorder
1 0.09507111
3 0.08239786
4 0.09345317
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.09395176
3 0.08385171
4 0.09229920
psychotic_disordersPsychotic disorder
1 0.1921904
3 0.1727261
4 0.2178664
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.1274059 0.1391611
3 0.1236762 0.1230710
4 0.1372178 0.1401488
Residual Deviance: 29549.19
AIC: 29729.19 create z and p values
make Z
z_clustered.phq <- summary(phq.model_clustered)$coefficients/summary(phq.model_clustered)$standard.errors
z_as.is.phq <- summary(phq.model)$coefficients/summary(phq.model)$standard.errorsprint z clusted ethnicity model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 -30.70745 9.085084 7.256237 -3.344237 -2.780620 -7.742493 -3.877527
3 -29.94580 3.899499 4.160747 -3.324993 -6.001187 -9.067317 -5.077441
4 -29.42255 5.835631 5.688466 -5.018682 -8.385554 -10.326855 -4.533106
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 -2.052753 -1.456128e+06 -0.3107757 2.369170
3 -4.090670 -2.045681e+00 2.8950826 2.699825
4 -3.684763 -1.845007e+00 4.1564134 3.034242
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 2.490967 4.630581
3 2.102137 1.964547
4 1.649335 3.513366
employment_prior_covidStudent employment_prior_covidUnemployed
1 3.6693996 18.10338
3 0.9210142 11.10095
4 1.5435391 11.80303
employment_changeDecreased employment employment_changeFurloughed
1 6.097038 4.390253
3 4.602005 4.501470
4 6.646903 3.764445
employment_changeIncreased employment key_workerGovernment defined key worker
1 1.0594727 3.331842
3 2.0101239 3.191447
4 0.2314938 2.645659
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 9.754026 16.80038
3 10.534048 19.71697
4 8.258779 16.59891
eating_disordersEating disorder
1 9.692795
3 7.069855
4 5.658961
obsessive_compulsive_disordersObsessive compulsive disorder
1 6.376396
3 2.578652
4 4.118626
psychotic_disordersPsychotic disorder
1 -0.5928962
3 -0.9973114
4 -2.4505819
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 8.416263 5.005661
3 3.834477 4.699070
4 3.315726 4.244750print z ethnicity as is
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 -30.70008 9.082330 7.260341 -3.327422 -2.744113 -7.716412 -3.870594
3 -29.90379 3.893371 4.153275 -3.305603 -5.994177 -9.053903 -5.068045
4 -29.34571 5.825472 5.690645 -5.036236 -8.411528 -10.352797 -4.549595
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 -2.047517 -6.985717e+06 -0.3030834 2.367790
3 -4.089330 -2.046267e+00 2.8668961 2.683645
4 -3.689879 -1.847706e+00 4.1502423 3.023720
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
1 2.2960627 1.3823566 0.3933091
3 2.1982830 0.5640188 0.4097952
4 0.8850066 0.3165557 0.7851257
ethnicityArab ethnicityOther employment_prior_covidRetired
1 -0.11457701 0.4043671 4.622216
3 0.72638972 0.3245459 1.957120
4 -0.08803967 1.6915639 3.519903
employment_prior_covidStudent employment_prior_covidUnemployed
1 3.6639837 18.10999
3 0.9164946 11.11388
4 1.5486490 11.77808
employment_changeDecreased employment employment_changeFurloughed
1 6.111861 4.393378
3 4.604172 4.494111
4 6.625886 3.742463
employment_changeIncreased employment key_workerGovernment defined key worker
1 1.0864317 3.313397
3 2.0248230 3.201204
4 0.1981475 2.645454
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 9.770635 16.78725
3 10.519155 19.70425
4 8.217813 16.56855
eating_disordersEating disorder
1 9.696265
3 7.081347
4 5.640016
obsessive_compulsive_disordersObsessive compulsive disorder
1 6.364407
3 2.569060
4 4.128065
psychotic_disordersPsychotic disorder
1 -0.5995952
3 -1.0011317
4 -2.4378744
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 8.419158 5.016848
3 3.836242 4.702097
4 3.317113 4.221445two tailed z test p values
p_clustered.phq <- (1 - pnorm(abs(z_clustered.phq), 0, 1)) * 2
p_as.is.phq <- (1 - pnorm(abs(z_as.is.phq), 0, 1)) * 2show p for two tailed z test for ethnicity clustered model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65
1 0 0.000000e+00 3.979039e-13 8.250917e-04 5.425514e-03 9.769963e-15
3 0 9.639175e-05 3.172090e-05 8.842073e-04 1.958799e-09 0.000000e+00
4 0 5.358735e-09 1.281860e-08 5.202710e-07 0.000000e+00 0.000000e+00
age66-70 age71-75 age76+ genderFemale
1 1.055237e-04 4.009657e-02 0.00000000 7.559711e-01
3 3.825528e-07 4.301282e-05 0.04078774 3.790588e-03
4 5.812267e-06 2.289157e-04 0.06503655 3.232826e-05
genderNon-binary/Prefer to self define
1 0.017828077
3 0.006937603
4 0.002411408
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.01273958 3.646404e-06
3 0.03554124 4.946666e-02
4 0.09907909 4.424671e-04
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.0002431208 0
3 0.3570430153 0
4 0.1227000329 0
employment_changeDecreased employment employment_changeFurloughed
1 1.080516e-09 1.132186e-05
3 4.184431e-06 6.748506e-06
4 2.993250e-11 1.669193e-04
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.28938457 0.0008627318
3 0.04441808 0.0014156211
4 0.81693124 0.0081532077
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.000000e+00 0
3 0.000000e+00 0
4 2.220446e-16 0
eating_disordersEating disorder
1 0.000000e+00
3 1.550982e-12
4 1.522920e-08
obsessive_compulsive_disordersObsessive compulsive disorder
1 1.813041e-10
3 9.918667e-03
4 3.811378e-05
psychotic_disordersPsychotic disorder
1 0.55325065
3 0.31861340
4 0.01426255
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.0000000000 5.567075e-07
3 0.0001258319 2.613485e-06
4 0.0009140554 2.188370e-05show p for two tailed z test for ethnicity as is model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65
1 0 0.000000e+00 3.861356e-13 8.765361e-04 6.067473e-03 1.199041e-14
3 0 9.886074e-05 3.277503e-05 9.477232e-04 2.045184e-09 0.000000e+00
4 0 5.695158e-09 1.265603e-08 4.747745e-07 0.000000e+00 0.000000e+00
age66-70 age71-75 age76+ genderFemale
1 1.085707e-04 4.060738e-02 0.00000000 7.618263e-01
3 4.019231e-07 4.326218e-05 0.04073007 4.145191e-03
4 5.374926e-06 2.243604e-04 0.06464496 3.321235e-05
genderNon-binary/Prefer to self define ethnicityMixed or multiple ethnicity
1 0.017894695 0.02167230
3 0.007282448 0.02792895
4 0.002496871 0.37615309
ethnicityAsian ethnicityBlack ethnicityArab ethnicityOther
1 0.1668622 0.6940912 0.9087804 0.68594278
3 0.5727413 0.6819562 0.4675999 0.74552479
4 0.7515808 0.4323799 0.9298451 0.09072915
employment_prior_covidRetired employment_prior_covidStudent
1 3.796619e-06 0.0002483226
3 5.033338e-02 0.3594075444
4 4.317043e-04 0.1214661249
employment_prior_covidUnemployed employment_changeDecreased employment
1 0 9.847620e-10
3 0 4.141100e-06
4 0 3.451728e-11
employment_changeFurloughed employment_changeIncreased employment
1 1.116030e-05 0.27728806
3 6.986125e-06 0.04288554
4 1.822257e-04 0.84292964
key_workerGovernment defined key worker anxiety_disordersAnxiety disorder
1 0.0009217019 0.000000e+00
3 0.0013685480 0.000000e+00
4 0.0081581445 2.220446e-16
depressive_disordersDepressive disorder eating_disordersEating disorder
1 0 0.000000e+00
3 0 1.427525e-12
4 0 1.700348e-08
obsessive_compulsive_disordersObsessive compulsive disorder
1 1.960461e-10
3 1.019747e-02
4 3.658293e-05
psychotic_disordersPsychotic disorder
1 0.54877604
3 0.31676312
4 0.01477391
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.0000000000 5.252597e-07
3 0.0001249311 2.575029e-06
4 0.0009095269 2.427412e-05Relative risk
Relative risk is the ratio of the probability of choosing one outcome category over the probability of choosing the baseline category.
Relative risk of 1 means there is no difference between outcome and baseline. Greater than 1 indicates outcome is more likely than baseline. Less than one indicates it is less likely than baseline.
RR 1.4 would mean there is a 1.4% chance of outcome occurring relative to baseline etc
if something is below one (e.g. 0.6 x the risk) you can also interpret this as 40% less probability of outcomes (1-0.6 = 0.4). So, if women are 0.6 x as likely to belong to a class as men, that means that there is a 40% lower probability of being in that category for women.
exponentiate the model coefficient to obtain relative risk for all predictors. Can see equations here
clustered ethnicity
relative_risk_clustered.phq <- data.frame(exp(coef(phq.model_clustered))) %>%
rownames_to_column() %>%
rename(class = "rowname",
intercept = "X.Intercept.") %>%
pivot_longer(!class) %>%
rename(predictor = name,
RelativeRisk = value)
relative_risk_clustered.phq$predictor <- gsub("76.","76+",relative_risk_clustered.phq$predictor)
relative_risk_clustered.phq ethnicity
relative_risk_as.is.phq <-data.frame(exp(coef(phq.model))) %>%
rownames_to_column() %>%
rename(class = "rowname",
intercept = "X.Intercept.") %>%
pivot_longer(!class) %>%
rename(predictor = name,
RelativeRisk = value)
relative_risk_as.is.phq$predictor <- gsub("76.","76+",relative_risk_as.is.phq$predictor)
head(relative_risk_as.is.phq)confidence intervals (rr scale)
The confint package has a specific method that is run for multinom objects (see methods(confint)) Obtain the exponent of upper and lower to get relative risk version
First create the CI (converted to relative risk) and reformat it so it is ready to merge with relative risk data for plotting
clustered ethnicity
CI_clustered.phq <- data.frame(exp(confint(phq.model_clustered,level=0.95))) %>%
rownames_to_column()
names(CI_clustered.phq) <- c("predictor",
"1_lower","1_upper",
"3_lower","3_upper",
"4_lower","4_upper"
)
CI_clustered.phq <- CI_clustered.phq %>%
pivot_longer(!predictor) %>%
separate(name, c("class","CI")) %>%
pivot_wider(names_from = CI,
values_from= value)
CI_clustered.phq$predictor <- gsub("-",".",CI_clustered.phq$predictor)
CI_clustered.phq$predictor <- gsub(" ",".",CI_clustered.phq$predictor)
CI_clustered.phq$predictor <- gsub("/",".",CI_clustered.phq$predictor)
CI_clustered.phq$predictor <- gsub("\\(Intercept\\)","intercept",CI_clustered.phq$predictor)
head(CI_clustered.phq)ethnicity as is
CI_as.is.phq<- data.frame(exp(confint(phq.model,level=0.95))) %>%
rownames_to_column()
names(CI_as.is.phq) <- c("predictor",
"1_lower","1_upper",
"3_lower","3_upper",
"4_lower","4_upper"
)
CI_as.is.phq <- CI_as.is.phq %>%
pivot_longer(!predictor) %>%
separate(name, c("class","CI")) %>%
pivot_wider(names_from = CI,
values_from= value)
CI_as.is.phq$predictor <- gsub("-",".",CI_as.is.phq$predictor)
CI_as.is.phq$predictor <- gsub(" ",".",CI_as.is.phq$predictor)
CI_as.is.phq$predictor <- gsub("/",".",CI_as.is.phq$predictor)
CI_as.is.phq$predictor <- gsub("\\(Intercept\\)","intercept",CI_as.is.phq$predictor)
head(CI_as.is.phq)forest plots of relative risk
PHQ
prepare and format plot data
plot_phq_clustered_dat$category <-plot_phq_clustered_dat$predictor
plot_phq_clustered_dat <-
plot_phq_clustered_dat %>%
mutate(category = case_when(grepl("age",predictor) ~ "Age",
grepl("gender",predictor) ~ "Gender",
grepl("ethnicity_clustered",predictor) ~ "Ethnicity",
grepl("employment_prior_covid",predictor) ~ "Employment before covid",
grepl("employment_change",predictor) ~ "Change in employment due to covid",
grepl("key_worker",predictor) ~ "Key worker status",
grepl("anxiety_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("depressive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("eating_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("obsessive_compulsive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("psychotic_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("personality_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("bipolar_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("intercept",predictor) ~ "Intercept"
),
predictor = case_when(grepl("age",predictor) ~ gsub("age","",predictor),
grepl("gender",predictor) ~ gsub("gender","",predictor),
grepl("ethnicity_clustered",predictor) ~ gsub("ethnicity_clustered","",predictor),
grepl("employment_prior_covid",predictor) ~ gsub("employment_prior_covid","",predictor),
grepl("employment_change",predictor) ~ gsub("employment_change","",predictor),
grepl("key_worker",predictor) ~ gsub("key_worker","",predictor),
grepl("anxiety_disorders",predictor) ~ gsub("anxiety_disorders","",predictor),
grepl("depressive_disorders",predictor) ~ gsub("depressive_disorders","",predictor),
grepl("eating_disorders",predictor) ~ gsub("eating_disorders","",predictor),
grepl("obsessive_compulsive_disorders",predictor) ~ gsub("obsessive_compulsive_disorders","",predictor),
grepl("psychotic_disorders",predictor) ~ gsub("psychotic_disorders","",predictor),
grepl("personality_disorders",predictor) ~ gsub("personality_disorders","",predictor),
grepl("bipolar_disorders",predictor) ~ gsub("bipolar_disorders","",predictor),
grepl("intercept",predictor) ~ "intercept")
)
plot_phq_as.is_dat$category <-plot_phq_as.is_dat$predictor
plot_phq_as.is_dat <-
plot_phq_as.is_dat %>%
mutate(category = case_when(grepl("age",predictor) ~ "Age",
grepl("gender",predictor) ~ "Gender",
grepl("ethnicity",predictor) ~ "Ethnicity",
grepl("employment_prior_covid",predictor) ~ "Employment before covid",
grepl("employment_change",predictor) ~ "Change in employment due to covid",
grepl("key_worker",predictor) ~ "Key worker status",
grepl("anxiety_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("depressive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("eating_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("obsessive_compulsive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("psychotic_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("personality_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("bipolar_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("intercept",predictor) ~ "Intercept"
),
predictor = case_when(grepl("age",predictor) ~ gsub("age","",predictor),
grepl("gender",predictor) ~ gsub("gender","",predictor),
grepl("ethnicity",predictor) ~ gsub("ethnicity_clustered","",predictor),
grepl("employment_prior_covid",predictor) ~ gsub("employment_prior_covid","",predictor),
grepl("employment_change",predictor) ~ gsub("employment_change","",predictor),
grepl("key_worker",predictor) ~ gsub("key_worker","",predictor),
grepl("anxiety_disorders",predictor) ~ gsub("anxiety_disorders","",predictor),
grepl("depressive_disorders",predictor) ~ gsub("depressive_disorders","",predictor),
grepl("eating_disorders",predictor) ~ gsub("eating_disorders","",predictor),
grepl("obsessive_compulsive_disorders",predictor) ~ gsub("obsessive_compulsive_disorders","",predictor),
grepl("psychotic_disorders",predictor) ~ gsub("psychotic_disorders","",predictor),
grepl("personality_disorders",predictor) ~ gsub("personality_disorders","",predictor),
grepl("bipolar_disorders",predictor) ~ gsub("bipolar_disorders","",predictor),
grepl("intercept",predictor) ~ "intercept")
)sort variables so categories are together
plot_phq_clustered_dat$category <- factor(plot_phq_clustered_dat$category,
levels=c("Intercept","Gender","Ethnicity","Age","Employment before covid","Change in employment due to covid","Key worker status","Pre-existing mental health diagnosis"))
plot_phq_clustered_dat$predictor <- fct_reorder2(plot_phq_clustered_dat$predictor, plot_phq_clustered_dat$class, plot_phq_clustered_dat$category)
plot_phq_as.is_dat$category <- factor(plot_phq_as.is_dat$category,
levels=c("Intercept","Gender","Ethnicity","Age","Employment before covid","Change in employment due to covid","Key worker status","Pre-existing mental health diagnosis"))
plot_phq_as.is_dat$predictor <- fct_reorder2(plot_phq_as.is_dat$predictor, plot_phq_as.is_dat$class, plot_phq_as.is_dat$category)forest plot clustered ethnicity
Note: smaller category error bars extend beyond a reasonable axis. Wil need fixing for publication.
phq_eth_clustered_forest <-relrisk.forest(plot_phq_clustered_dat,
plot_title="Relative risk (95% CI) of class membership: depression",
sub_title="ethnicity clustered",
lowlim = 0.001,
uplim = 10)
phq_eth_clustered_forestWarning: Removed 1 rows containing missing values (geom_point).
forest plot ethnicity as is
Note: smaller category error bars extend beyond a reasonable axis. Wil need fixing for publication.
phq_eth_as.is_forest <- relrisk.forest(plot_phq_as.is_dat,
plot_title="Relative risk (95% CI) of class membership: depression",
sub_title="ethnicity as is",
lowlim = 0.001,
uplim = 10)
phq_eth_as.is_forestWarning: Removed 2 rows containing missing values (geom_point).
#### save plots
ggsave(
filename=file.path(dirname(dirname(getwd())),"output/ForestPlots/PHQ/depression_RelativeRisk_forestPlot_ethnicityClustered.png"),
plot=phq_eth_clustered_forest,
width = 18, height = 10, dpi = 150, units = "in")Warning: Removed 1 rows containing missing values (geom_point).ggsave(
filename=file.path(dirname(dirname(getwd())),"output/ForestPlots/PHQ/depression_RelativeRisk_forestPlot.png"),
plot=phq_eth_as.is_forest,
width = 18, height = 10, dpi = 150, units = "in")Warning: Removed 2 rows containing missing values (geom_point).GAD
prepare and run models
specify reference category for outcome
will use the largest class group as reference
BiggestClass.gad <- gad.AllData %>%
group_by(MostLikelyClass) %>%
summarise(n=n()) %>%
arrange(-n) %>%
slice(1)%>%
select(MostLikelyClass) %>%
as.character(.)
BiggestClass.gad[1] "1"Will need to adapt this to the correct number of classes depending on final model
Perform regression (ethnicity clustered)
# weights: 108 (78 variable)
initial value 22478.763066
iter 10 value 14824.087869
iter 20 value 13416.842887
iter 30 value 13142.479298
iter 40 value 13071.880866
iter 50 value 13049.903786
iter 60 value 13034.906346
iter 70 value 13032.108909
iter 80 value 13031.900653
iter 90 value 13031.859302
final value 13031.857624
convergedPerform regression (ethnicity as is)
# weights: 124 (90 variable)
initial value 22478.763066
iter 10 value 14813.923252
iter 20 value 13422.343932
iter 30 value 13137.531913
iter 40 value 13064.638921
iter 50 value 13042.619448
iter 60 value 13027.215094
iter 70 value 13024.194017
iter 80 value 13024.022070
iter 90 value 13023.887259
iter 100 value 13023.867221
final value 13023.867221
stopped after 100 iterationsView and process results
1: coefficients 2: std errors 3: residual deviances & AIC ##### view summary: Ethnicity clustered
Call:
multinom(formula = formula_clustered, data = gad.AllData)
Coefficients:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
2 -3.650424 1.1671214 0.5167330 -0.3085380 -0.343064521 -0.7999582 -2.365594
3 -3.475616 1.0496975 0.3779647 -0.2286434 -0.534428182 -0.7652060 -1.189222
4 -3.673601 0.1425086 0.3313403 0.1073131 -0.005039378 -0.4684441 -0.753368
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
2 -0.9020561 -1.4383121 0.2448057 0.6393261
3 -1.4549421 0.5644221 0.4638501 0.2542844
4 -0.3070324 -10.1905261 0.3037216 0.5180137
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
2 0.0592478 1.0102689
3 0.2659093 -0.3294263
4 0.1400527 0.4950650
employment_prior_covidStudent employment_prior_covidUnemployed
2 1.2912962 1.1014962
3 0.5522132 0.3372711
4 0.4724846 0.3397331
employment_changeDecreased employment employment_changeFurloughed
2 0.4543803 0.2636938
3 0.2660162 0.1221211
4 0.1787511 0.2495741
employment_changeIncreased employment key_workerGovernment defined key worker
2 0.30259582 0.1241316
3 0.06364632 0.1524750
4 -0.01003740 0.1400552
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 1.341481 0.8639027
3 1.017514 0.5559895
4 1.014918 0.6965658
eating_disordersEating disorder
2 0.6927949
3 0.4714443
4 0.3468174
obsessive_compulsive_disordersObsessive compulsive disorder
2 0.5245066
3 0.3518988
4 0.3286507
psychotic_disordersPsychotic disorder
2 -0.27256668
3 -0.37148039
4 -0.01097946
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 0.7749959 0.3766943
3 0.3466961 0.3097247
4 0.4947044 0.3160350
Std. Errors:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
2 0.1115779 0.1821392 0.08662416 0.07599623 0.07379979 0.09749191 0.5171385
3 0.1209820 0.2100379 0.10223948 0.08583027 0.08811655 0.11020444 0.3503438
4 0.1212064 0.3056786 0.11251066 0.08751113 0.08621793 0.11082011 0.3118558
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
2 0.4102723 1.034236e+00 0.07006872 0.1955054
3 0.5916734 4.990777e-01 0.08497578 0.2801124
4 0.3794280 5.425413e-06 0.07958294 0.2547380
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
2 0.1296068 0.3478351
3 0.1405756 0.6066659
4 0.1492563 0.3994340
employment_prior_covidStudent employment_prior_covidUnemployed
2 0.4284839 0.08869572
3 0.5929406 0.11846162
4 0.5930770 0.11775712
employment_changeDecreased employment employment_changeFurloughed
2 0.07550160 0.07667607
3 0.08745712 0.08780095
4 0.08980712 0.08564925
employment_changeIncreased employment key_workerGovernment defined key worker
2 0.1025110 0.06079071
3 0.1267405 0.06833989
4 0.1318353 0.06847648
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 0.07929849 0.08052156
3 0.08484729 0.08492792
4 0.08406461 0.08673466
eating_disordersEating disorder
2 0.07895395
3 0.09658515
4 0.10152638
obsessive_compulsive_disordersObsessive compulsive disorder
2 0.07843148
3 0.09753485
4 0.09930082
psychotic_disordersPsychotic disorder
2 0.1687666
3 0.2298316
4 0.2003119
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 0.1052235 0.1170634
3 0.1400174 0.1485819
4 0.1360787 0.1438603
Residual Deviance: 26063.72
AIC: 26219.72 view summary: Ethnicity as is
Call:
multinom(formula = formula, data = gad.AllData)
Coefficients:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
2 -3.646697 1.1667348 0.5156172 -0.3110604 -0.34601962 -0.8031963 -2.3728387
3 -3.469596 1.0509495 0.3793907 -0.2312247 -0.53736659 -0.7691526 -1.1995273
4 -3.686265 0.1473397 0.3303717 0.1070361 -0.00347322 -0.4660158 -0.7482131
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
2 -0.9040785 -1.4422427 0.2447487 0.6397616
3 -1.4554868 0.5597647 0.4647165 0.2525852
4 -0.3024041 -8.7921182 0.3078017 0.5318378
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
2 -0.05176409 -0.0003202253 -0.1430201
3 0.12618985 0.1578156311 0.3250710
4 -0.07803545 0.5743700321 0.6327083
ethnicityArab ethnicityOther employment_prior_covidRetired
2 1.322949 0.4771204 1.0146878
3 -6.923904 0.7987081 -0.3221631
4 -6.750052 -0.5325362 0.5006125
employment_prior_covidStudent employment_prior_covidUnemployed
2 1.2991101 1.1010880
3 0.5649541 0.3347179
4 0.4802050 0.3397831
employment_changeDecreased employment employment_changeFurloughed
2 0.4524217 0.2638354
3 0.2650956 0.1206205
4 0.1798413 0.2530216
employment_changeIncreased employment key_workerGovernment defined key worker
2 0.30181772 0.1250909
3 0.05921025 0.1526558
4 -0.01028472 0.1390436
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 1.338478 0.8645838
3 1.014348 0.5540216
4 1.019877 0.7019997
eating_disordersEating disorder
2 0.6921110
3 0.4679514
4 0.3450005
obsessive_compulsive_disordersObsessive compulsive disorder
2 0.5257258
3 0.3536987
4 0.3306319
psychotic_disordersPsychotic disorder
2 -0.26935131
3 -0.36799398
4 -0.01354605
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 0.7743597 0.3760422
3 0.3471852 0.3058703
4 0.4947032 0.3173593
Std. Errors:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
2 0.1116937 0.1822423 0.08665573 0.07603363 0.07384173 0.0975301 0.5173850
3 0.1211115 0.2101064 0.10226111 0.08587586 0.08816022 0.1102456 0.3507171
4 0.1214361 0.3057609 0.11255457 0.08755993 0.08626089 0.1108650 0.3119076
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
2 0.4101659 1.034099 0.07010760 0.1955637
3 0.5907168 0.498965 0.08504013 0.2803780
4 0.3795802 50.638438 0.07965851 0.2548989
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
2 0.1783553 0.2787059 0.4671697
3 0.1969998 0.2953030 0.4636567
4 0.2179014 0.2547151 0.4131507
ethnicityArab ethnicityOther employment_prior_covidRetired
2 1.425252 0.3005387 0.3477641
3 72.668274 0.3211487 0.6060899
4 69.796153 0.5325911 0.3993792
employment_prior_covidStudent employment_prior_covidUnemployed
2 0.4286759 0.08874296
3 0.5920381 0.11854461
4 0.5934081 0.11780540
employment_changeDecreased employment employment_changeFurloughed
2 0.07555035 0.07670187
3 0.08748728 0.08783670
4 0.08984714 0.08571794
employment_changeIncreased employment key_workerGovernment defined key worker
2 0.1025941 0.06083141
3 0.1268516 0.06837122
4 0.1319701 0.06849874
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 0.07934682 0.08058216
3 0.08491949 0.08497846
4 0.08418799 0.08681954
eating_disordersEating disorder
2 0.07895700
3 0.09660906
4 0.10161878
obsessive_compulsive_disordersObsessive compulsive disorder
2 0.07843933
3 0.09755387
4 0.09935671
psychotic_disordersPsychotic disorder
2 0.1687378
3 0.2298358
4 0.2004391
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 0.1052320 0.1171095
3 0.1400425 0.1486702
4 0.1360719 0.1439566
Residual Deviance: 26047.73
AIC: 26227.73 create z and p values
make Z
z_clustered.gad <- summary(gad.model_clustered)$coefficients/summary(gad.model_clustered)$standard.errors
z_as.is.gad <- summary(gad.model)$coefficients/summary(gad.model)$standard.errorsprint z clusted ethnicity model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
2 -32.71636 6.407855 5.965230 -4.059912 -4.6485841 -8.205380 -4.574390
3 -28.72836 4.997658 3.696856 -2.663902 -6.0650148 -6.943513 -3.394444
4 -30.30865 0.466204 2.944968 1.226280 -0.0584493 -4.227068 -2.415758
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
2 -2.1986765 -1.390700e+00 3.493794 3.2701194
3 -2.4590291 1.130930e+00 5.458616 0.9077944
4 -0.8091979 -1.878295e+06 3.816416 2.0335153
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
2 0.4571349 2.904448
3 1.8915760 -0.543011
4 0.9383370 1.239416
employment_prior_covidStudent employment_prior_covidUnemployed
2 3.0136401 12.418819
3 0.9313128 2.847091
4 0.7966666 2.885032
employment_changeDecreased employment employment_changeFurloughed
2 6.018154 3.439063
3 3.041675 1.390886
4 1.990389 2.913908
employment_changeIncreased employment key_workerGovernment defined key worker
2 2.95183704 2.041951
3 0.50217813 2.231128
4 -0.07613593 2.045303
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 16.91685 10.728838
3 11.99229 6.546605
4 12.07307 8.030998
eating_disordersEating disorder
2 8.774671
3 4.881126
4 3.416032
obsessive_compulsive_disordersObsessive compulsive disorder
2 6.687450
3 3.607929
4 3.309647
psychotic_disordersPsychotic disorder
2 -1.61505149
3 -1.61631565
4 -0.05481183
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 7.365239 3.217865
3 2.476093 2.084539
4 3.635428 2.196818print z ethnicity as is
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
2 -32.64908 6.402108 5.950181 -4.091089 -4.68596312 -8.235368 -4.586215
3 -28.64795 5.001987 3.710020 -2.692545 -6.09534096 -6.976716 -3.420213
4 -30.35559 0.481879 2.935213 1.222433 -0.04026414 -4.203454 -2.398830
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
2 -2.2041775 -1.3946859 3.491043 3.2713725
3 -2.4639334 1.1218518 5.464674 0.9008739
4 -0.7966804 -0.1736254 3.864015 2.0864655
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
2 -0.2902302 -0.001148972 -0.3061417
3 0.6405584 0.534419244 0.7011028
4 -0.3581228 2.254951039 1.5314226
ethnicityArab ethnicityOther employment_prior_covidRetired
2 0.92822109 1.587550 2.9177471
3 -0.09528097 2.487035 -0.5315435
4 -0.09671094 -0.999897 1.2534765
employment_prior_covidStudent employment_prior_covidUnemployed
2 3.0305182 12.407610
3 0.9542529 2.823561
4 0.8092322 2.884274
employment_changeDecreased employment employment_changeFurloughed
2 5.988347 3.439752
3 3.030104 1.373236
4 2.001637 2.951793
employment_changeIncreased employment key_workerGovernment defined key worker
2 2.94186230 2.056354
3 0.46676803 2.232750
4 -0.07793218 2.029870
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 16.86871 10.729220
3 11.94482 6.519554
4 12.11428 8.085734
eating_disordersEating disorder
2 8.765671
3 4.843763
4 3.395046
obsessive_compulsive_disordersObsessive compulsive disorder
2 6.702325
3 3.625676
4 3.327726
psychotic_disordersPsychotic disorder
2 -1.59627176
3 -1.60111702
4 -0.06758187
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 7.358595 3.211031
3 2.479142 2.057374
4 3.635602 2.204548two tailed z test p values
p_clustered.gad <- (1 - pnorm(abs(z_clustered.gad), 0, 1)) * 2
p_as.is.gad <- (1 - pnorm(abs(z_as.is.gad), 0, 1)) * 2show p for two tailed z test for ethnicity clustered model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65
2 0 1.475813e-10 2.442897e-09 4.909115e-05 3.342214e-06 2.220446e-16
3 0 5.803079e-07 2.182857e-04 7.724014e-03 1.319415e-09 3.824718e-12
4 0 6.410695e-01 3.229883e-03 2.200935e-01 9.533907e-01 2.367561e-05
age66-70 age71-75 age76+ genderFemale
2 4.776088e-06 0.02790093 0.1643165 4.762076e-04
3 6.876817e-04 0.01393133 0.2580844 4.798605e-08
4 1.570251e-02 0.41840135 0.0000000 1.354044e-04
genderNon-binary/Prefer to self define
2 0.001075021
3 0.363986836
4 0.042000490
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
2 0.64757408 0.003679016
3 0.05854749 0.587122199
4 0.34807126 0.215191371
employment_prior_covidStudent employment_prior_covidUnemployed
2 0.002581339 0.000000000
3 0.351691757 0.004412069
4 0.425644713 0.003913733
employment_changeDecreased employment employment_changeFurloughed
2 1.764171e-09 0.0005837308
3 2.352654e-03 0.1642599131
4 4.654812e-02 0.0035693475
employment_changeIncreased employment key_workerGovernment defined key worker
2 0.003158896 0.04115641
3 0.615542229 0.02567268
4 0.939310952 0.04082496
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
2 0 0.000000e+00
3 0 5.885981e-11
4 0 8.881784e-16
eating_disordersEating disorder
2 0.000000e+00
3 1.054820e-06
4 6.354079e-04
obsessive_compulsive_disordersObsessive compulsive disorder
2 2.270939e-11
3 3.086507e-04
4 9.341355e-04
psychotic_disordersPsychotic disorder
2 0.1062995
3 0.1060261
4 0.9562884
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 1.767475e-13 0.001291487
3 1.328290e-02 0.037111202
4 2.775191e-04 0.028033458show p for two tailed z test for ethnicity as is model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65
2 0 1.532459e-10 2.678469e-09 4.293519e-05 2.786464e-06 2.220446e-16
3 0 5.674237e-07 2.072432e-04 7.090886e-03 1.092044e-09 3.021583e-12
4 0 6.298919e-01 3.333184e-03 2.215439e-01 9.678825e-01 2.628730e-05
age66-70 age71-75 age76+ genderFemale
2 4.513545e-06 0.02751186 0.1631106 4.811381e-04
3 6.257205e-04 0.01374216 0.2619255 4.637595e-08
4 1.644757e-02 0.42563666 0.8621599 1.115383e-04
genderNon-binary/Prefer to self define ethnicityMixed or multiple ethnicity
2 0.001070268 0.7716402
3 0.367655376 0.5218096
4 0.036936476 0.7202514
ethnicityAsian ethnicityBlack ethnicityArab ethnicityOther
2 0.9990833 0.7594967 0.3532929 0.11238805
3 0.5930515 0.4832388 0.9240917 0.01288128
4 0.0241364 0.1256650 0.9229559 0.31736037
employment_prior_covidRetired employment_prior_covidStudent
2 0.003525701 0.002441345
3 0.595042224 0.339955648
4 0.210032339 0.418381579
employment_prior_covidUnemployed employment_changeDecreased employment
2 0.000000000 2.119847e-09
3 0.004749336 2.444696e-03
4 0.003923167 4.532384e-02
employment_changeFurloughed employment_changeIncreased employment
2 0.0005822465 0.00326245
3 0.1696790547 0.64066585
4 0.0031593431 0.93788200
key_workerGovernment defined key worker anxiety_disordersAnxiety disorder
2 0.03974843 0
3 0.02556546 0
4 0.04236972 0
depressive_disordersDepressive disorder eating_disordersEating disorder
2 0.000000e+00 0.000000e+00
3 7.051693e-11 1.274027e-06
4 6.661338e-16 6.861704e-04
obsessive_compulsive_disordersObsessive compulsive disorder
2 2.051292e-11
3 2.882067e-04
4 8.755800e-04
psychotic_disordersPsychotic disorder
2 0.1104281
3 0.1093510
4 0.9461185
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
2 1.858513e-13 0.001322596
3 1.316989e-02 0.039650249
4 2.773318e-04 0.027485825Relative risk
Relative risk is the ratio of the probability of choosing one outcome category over the probability of choosing the baseline category.
Relative risk of 1 means there is no difference between outcome and baseline. Greater than 1 indicates outcome is more likely than baseline. Less than one indicates it is less likely than baseline.
RR 1.4 would mean there is a 1.4% chance of outcome occurring relative to baseline etc
if something is below one (e.g. 0.6 x the risk) you can also interpret this as 40% less probability of outcomes (1-0.6 = 0.4). So, if women are 0.6 x as likely to belong to a class as men, that means that there is a 40% lower probability of being in that category for women.
exponentiate the model coefficient to obtain relative risk for all predictors. Can see equations here
clustered ethnicity
relative_risk_clustered.gad <- data.frame(exp(coef(gad.model_clustered))) %>%
rownames_to_column() %>%
rename(class = "rowname",
intercept = "X.Intercept.") %>%
pivot_longer(!class) %>%
rename(predictor = name,
RelativeRisk = value)
relative_risk_clustered.gad$predictor <- gsub("76.","76+",relative_risk_clustered.gad$predictor)
relative_risk_clustered.gad ethnicity
relative_risk_as.is.gad <-data.frame(exp(coef(gad.model))) %>%
rownames_to_column() %>%
rename(class = "rowname",
intercept = "X.Intercept.") %>%
pivot_longer(!class) %>%
rename(predictor = name,
RelativeRisk = value)
relative_risk_as.is.gad$predictor <- gsub("76.","76+",relative_risk_as.is.gad$predictor)
head(relative_risk_as.is.gad)confidence intervals (rr scale)
The confint package has a specific method that is run for multinom objects (see methods(confint)) Obtain the exponent of upper and lower to get relative risk version
First create the CI (converted to relative risk) and reformat it so it is ready to merge with relative risk data for plotting
clustered ethnicity
CI_clustered.gad <- data.frame(exp(confint(gad.model_clustered,level=0.95))) %>%
rownames_to_column()
names(CI_clustered.gad) <- c("predictor",
"2_lower","2_upper",
"3_lower","3_upper",
"4_lower","4_upper"
)
CI_clustered.gad <- CI_clustered.gad %>%
pivot_longer(!predictor) %>%
separate(name, c("class","CI")) %>%
pivot_wider(names_from = CI,
values_from= value)
CI_clustered.gad$predictor <- gsub("-",".",CI_clustered.gad$predictor)
CI_clustered.gad$predictor <- gsub(" ",".",CI_clustered.gad$predictor)
CI_clustered.gad$predictor <- gsub("/",".",CI_clustered.gad$predictor)
CI_clustered.gad$predictor <- gsub("\\(Intercept\\)","intercept",CI_clustered.gad$predictor)
head(CI_clustered.gad)ethnicity as is
CI_as.is.gad<- data.frame(exp(confint(gad.model,level=0.95))) %>%
rownames_to_column()
names(CI_as.is.gad) <- c("predictor",
"2_lower","2_upper",
"3_lower","3_upper",
"4_lower","4_upper"
)
CI_as.is.gad <- CI_as.is.gad %>%
pivot_longer(!predictor) %>%
separate(name, c("class","CI")) %>%
pivot_wider(names_from = CI,
values_from= value)
CI_as.is.gad$predictor <- gsub("-",".",CI_as.is.gad$predictor)
CI_as.is.gad$predictor <- gsub(" ",".",CI_as.is.gad$predictor)
CI_as.is.gad$predictor <- gsub("/",".",CI_as.is.gad$predictor)
CI_as.is.gad$predictor <- gsub("\\(Intercept\\)","intercept",CI_as.is.gad$predictor)
head(CI_as.is.gad)forest plots of relative risk
GAD
prepare and format plot data
plot_gad_clustered_dat$category <-plot_gad_clustered_dat$predictor
plot_gad_clustered_dat <-
plot_gad_clustered_dat %>%
mutate(category = case_when(grepl("age",predictor) ~ "Age",
grepl("gender",predictor) ~ "Gender",
grepl("ethnicity_clustered",predictor) ~ "Ethnicity",
grepl("employment_prior_covid",predictor) ~ "Employment before covid",
grepl("employment_change",predictor) ~ "Change in employment due to covid",
grepl("key_worker",predictor) ~ "Key worker status",
grepl("anxiety_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("depressive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("eating_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("obsessive_compulsive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("psychotic_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("personality_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("bipolar_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("intercept",predictor) ~ "Intercept"
),
predictor = case_when(grepl("age",predictor) ~ gsub("age","",predictor),
grepl("gender",predictor) ~ gsub("gender","",predictor),
grepl("ethnicity_clustered",predictor) ~ gsub("ethnicity_clustered","",predictor),
grepl("employment_prior_covid",predictor) ~ gsub("employment_prior_covid","",predictor),
grepl("employment_change",predictor) ~ gsub("employment_change","",predictor),
grepl("key_worker",predictor) ~ gsub("key_worker","",predictor),
grepl("anxiety_disorders",predictor) ~ gsub("anxiety_disorders","",predictor),
grepl("depressive_disorders",predictor) ~ gsub("depressive_disorders","",predictor),
grepl("eating_disorders",predictor) ~ gsub("eating_disorders","",predictor),
grepl("obsessive_compulsive_disorders",predictor) ~ gsub("obsessive_compulsive_disorders","",predictor),
grepl("psychotic_disorders",predictor) ~ gsub("psychotic_disorders","",predictor),
grepl("personality_disorders",predictor) ~ gsub("personality_disorders","",predictor),
grepl("bipolar_disorders",predictor) ~ gsub("bipolar_disorders","",predictor),
grepl("intercept",predictor) ~ "intercept")
)
plot_gad_as.is_dat$category <-plot_gad_as.is_dat$predictor
plot_gad_as.is_dat <-
plot_gad_as.is_dat %>%
mutate(category = case_when(grepl("age",predictor) ~ "Age",
grepl("gender",predictor) ~ "Gender",
grepl("ethnicity",predictor) ~ "Ethnicity",
grepl("employment_prior_covid",predictor) ~ "Employment before covid",
grepl("employment_change",predictor) ~ "Change in employment due to covid",
grepl("key_worker",predictor) ~ "Key worker status",
grepl("anxiety_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("depressive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("eating_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("obsessive_compulsive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("psychotic_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("personality_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("bipolar_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("intercept",predictor) ~ "Intercept"
),
predictor = case_when(grepl("age",predictor) ~ gsub("age","",predictor),
grepl("gender",predictor) ~ gsub("gender","",predictor),
grepl("ethnicity",predictor) ~ gsub("ethnicity_clustered","",predictor),
grepl("employment_prior_covid",predictor) ~ gsub("employment_prior_covid","",predictor),
grepl("employment_change",predictor) ~ gsub("employment_change","",predictor),
grepl("key_worker",predictor) ~ gsub("key_worker","",predictor),
grepl("anxiety_disorders",predictor) ~ gsub("anxiety_disorders","",predictor),
grepl("depressive_disorders",predictor) ~ gsub("depressive_disorders","",predictor),
grepl("eating_disorders",predictor) ~ gsub("eating_disorders","",predictor),
grepl("obsessive_compulsive_disorders",predictor) ~ gsub("obsessive_compulsive_disorders","",predictor),
grepl("psychotic_disorders",predictor) ~ gsub("psychotic_disorders","",predictor),
grepl("personality_disorders",predictor) ~ gsub("personality_disorders","",predictor),
grepl("bipolar_disorders",predictor) ~ gsub("mental_health_controls","",predictor),
grepl("intercept",predictor) ~ "intercept")
)sort variables so categories are together
plot_gad_clustered_dat$category <- factor(plot_gad_clustered_dat$category,
levels=c("Intercept","Gender","Ethnicity","Age","Employment before covid","Change in employment due to covid","Key worker status","Pre-existing mental health diagnosis"))
plot_gad_clustered_dat$predictor <- fct_reorder2(plot_gad_clustered_dat$predictor, plot_gad_clustered_dat$class, plot_gad_clustered_dat$category)
plot_gad_as.is_dat$category <- factor(plot_gad_as.is_dat$category,
levels=c("Intercept","Gender","Ethnicity","Age","Employment before covid","Change in employment due to covid","Key worker status","Pre-existing mental health diagnosis"))
plot_gad_as.is_dat$predictor <- fct_reorder2(plot_gad_as.is_dat$predictor, plot_gad_as.is_dat$class, plot_gad_as.is_dat$category)forest plot clustered ethnicity
Note: smaller category error bars extend beyond a reasonable axis. Wil need fixing for publication.
gad_eth_clustered_forest <-relrisk.forest(plot_gad_clustered_dat,
plot_title="Relative risk (95% CI) of class membership: anxiety",
sub_title="ethnicity clustered",
lowlim = 0.001,
uplim = 10)
gad_eth_clustered_forestWarning: Removed 1 rows containing missing values (geom_point).
forest plot ethnicity as is
Note: smaller category error bars extend beyond a reasonable axis. Wil need fixing for publication.
gad_eth_as.is_forest <- relrisk.forest(plot_gad_as.is_dat,
plot_title="Relative risk (95% CI) of class membership: anxiety",
sub_title="ethnicity as is",
lowlim = 0.001,
uplim = 10)
gad_eth_as.is_forestWarning: Removed 2 rows containing missing values (geom_point).
#### save plots
ggsave(
filename=file.path(dirname(dirname(getwd())),"output/ForestPlots/GAD/anxiety_RelativeRisk_forestPlot_ethnicityClustered.png"),
plot=gad_eth_clustered_forest,
width = 18, height = 10, dpi = 150, units = "in")Warning: Removed 1 rows containing missing values (geom_point).ggsave(
filename=file.path(dirname(dirname(getwd())),"output/ForestPlots/GAD/anxiety_RelativeRisk_forestPlot.png"),
plot=gad_eth_as.is_forest,
width = 18, height = 10, dpi = 150, units = "in")Warning: Removed 2 rows containing missing values (geom_point).MASQ
prepare and run models
specify reference category for outcome
will use the largest class group as reference
BiggestClass.masq <- masq.AllData %>%
group_by(MostLikelyClass) %>%
summarise(n=n()) %>%
arrange(-n) %>%
slice(1)%>%
select(MostLikelyClass) %>%
as.character(.)
BiggestClass.masq[1] "3"Will need to adapt this to the correct number of classes depending on final model
Perform regression (ethnicity clustered)
# weights: 135 (104 variable)
initial value 22086.316472
iter 10 value 12682.548064
iter 20 value 11859.394528
iter 30 value 11165.449558
iter 40 value 10972.460273
iter 50 value 10953.018144
iter 60 value 10949.412688
iter 70 value 10948.984909
iter 80 value 10948.782059
iter 90 value 10948.621659
iter 100 value 10948.568540
final value 10948.568540
stopped after 100 iterationsPerform regression (ethnicity as is)
# weights: 155 (120 variable)
initial value 22086.316472
iter 10 value 12681.333588
iter 20 value 11856.224058
iter 30 value 11165.483136
iter 40 value 10967.798201
iter 50 value 10947.912165
iter 60 value 10944.405467
iter 70 value 10943.778566
iter 80 value 10943.480845
iter 90 value 10943.189106
iter 100 value 10943.069793
final value 10943.069793
stopped after 100 iterationsView and process results
1: coefficients 2: std errors 3: residual deviances & AIC ##### view summary: Ethnicity clustered
Call:
multinom(formula = formula_clustered, data = masq.AllData)
Coefficients:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65
1 -3.3534339 -0.2246741 -0.27010497 -0.0480830 -0.44509283 -0.4155389
2 0.1708474 -0.3079949 0.07977917 -0.1449227 -0.25211813 -0.1401873
4 -3.4052826 -0.7173888 0.11844416 -0.1225591 -0.02773374 -0.2478546
5 -3.7150027 0.3529555 -0.08392221 -0.3424730 -0.33277432 -0.2616829
age66-70 age71-75 age76+ genderFemale
1 -0.53716100 -0.56984866 -5.9507835 0.1569743
2 0.09093635 -0.03069305 0.5692933 -0.4114874
4 -0.70999920 -0.36977132 -5.8712138 0.5539187
5 -0.86858283 -0.67315625 1.7192703 0.4443630
genderNon-binary/Prefer to self define
1 -0.06891821
2 -0.29824258
4 -8.79013616
5 0.53966418
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.09033541 -0.6183422
2 0.21639694 -0.4655660
4 0.17409120 -0.6511315
5 0.21466303 -0.7292526
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.8283103 -0.5044174
2 -0.7928715 -0.8130652
4 -6.0582890 -0.8721267
5 0.9471081 -0.3800778
employment_changeDecreased employment employment_changeFurloughed
1 -0.02202836 -0.16141492
2 -0.14918425 0.01364417
4 0.28898275 0.24859957
5 0.55499016 0.40275886
employment_changeIncreased employment key_workerGovernment defined key worker
1 -0.28851977 -0.12223381
2 0.05326513 0.06049872
4 0.01206147 0.09970771
5 0.27626513 0.00561951
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 -0.04550084 0.1915367
2 -0.45641233 -0.9242325
4 -0.01059467 -0.5770875
5 -0.23254922 0.1814364
eating_disordersEating disorder
1 0.02221847
2 -0.35745658
4 -0.09673019
5 0.11384672
obsessive_compulsive_disordersObsessive compulsive disorder
1 -0.49238729
2 -0.01646093
4 -0.07723958
5 -0.70983002
psychotic_disordersPsychotic disorder
1 -1.1022739
2 0.3006774
4 1.0650131
5 0.3958391
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 -0.1030690 0.4655088
2 -0.6311304 -0.4095350
4 -0.5666792 -0.4232159
5 -0.0612022 0.1286832
Std. Errors:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 0.22793179 0.6044032 0.26424974 0.17419140 0.18202038 0.20970425 0.5344713
2 0.07231369 0.2523324 0.09883712 0.06933067 0.06541695 0.06937663 0.1287773
4 0.21660033 0.7301498 0.23101953 0.17745715 0.16409047 0.19202904 0.5300329
5 0.23057104 0.4156004 0.23094755 0.18222530 0.17166663 0.19391529 0.6023694
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 0.7356971 24.6320715 0.16184059 0.6101184
2 0.1702867 0.3317533 0.04685269 0.2310442
4 0.6095128 24.4099826 0.15734105 53.6686568
5 0.7317874 0.6572730 0.16710814 0.4910483
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.3145273 1.0338619
2 0.1108999 0.2397209
4 0.2899838 1.0302742
5 0.2917672 1.0420769
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.7549094 0.2643221
2 0.7501142 0.1118174
4 25.5920219 0.3273417
5 0.7570266 0.2643364
employment_changeDecreased employment employment_changeFurloughed
1 0.17899055 0.1882251
2 0.06403244 0.0611963
4 0.15679799 0.1581659
5 0.15693288 0.1657373
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.30370305 0.13615317
2 0.09680611 0.04581351
4 0.26006389 0.12367543
5 0.25011468 0.13064332
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.16027616 0.1680468
2 0.05564113 0.0535519
4 0.14619694 0.1433618
5 0.15128673 0.1574589
eating_disordersEating disorder
1 0.2164463
2 0.1009802
4 0.2185169
5 0.1992213
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.26327149
2 0.09771346
4 0.23141461
5 0.26974454
psychotic_disordersPsychotic disorder
1 0.7281099
2 0.2010672
4 0.3595626
5 0.3902274
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.3059850 0.2834880
2 0.1721049 0.1649833
4 0.3804970 0.3885685
5 0.2897254 0.2971946
Residual Deviance: 21897.14
AIC: 22105.14 view summary: Ethnicity as is
Call:
multinom(formula = formula, data = masq.AllData)
Coefficients:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65
1 -3.3532860 -0.2186131 -0.26837948 -0.04825197 -0.44422205 -0.4145852
2 0.1693906 -0.3054205 0.08206832 -0.14505541 -0.25111543 -0.1394304
4 -3.4157541 -0.7300027 0.11756126 -0.12474580 -0.02557965 -0.2450459
5 -3.7277774 0.3376130 -0.08507963 -0.34766775 -0.32893789 -0.2540240
age66-70 age71-75 age76+ genderFemale
1 -0.53481212 -0.57099237 -6.4180629 0.1590440
2 0.09195194 -0.02995838 0.5719441 -0.4113503
4 -0.70591381 -0.36205067 -6.3226809 0.5548626
5 -0.88268331 -0.66980771 1.7208891 0.4498306
genderNon-binary/Prefer to self define ethnicityMixed or multiple ethnicity
1 -0.05986941 0.01169794
2 -0.30056502 0.25431192
4 -9.87255482 0.29701758
5 0.55041955 -0.02701734
ethnicityAsian ethnicityBlack ethnicityArab ethnicityOther
1 0.1473725 0.001870045 -2.859050 0.30483369
2 0.2648712 0.149374753 -6.025050 0.05759361
4 0.5142488 -0.237826749 -3.934510 -8.14998843
5 0.8821718 -7.602556799 -3.199124 0.21387162
employment_prior_covidRetired employment_prior_covidStudent
1 -0.6161812 0.8267753
2 -0.4694438 -0.7942560
4 -0.6532973 -6.5678224
5 -0.7278842 0.9554902
employment_prior_covidUnemployed employment_changeDecreased employment
1 -0.5077002 -0.02490025
2 -0.8139698 -0.14776472
4 -0.8637742 0.29550063
5 -0.3812142 0.55818983
employment_changeFurloughed employment_changeIncreased employment
1 -0.16366287 -0.29223232
2 0.01388549 0.05460684
4 0.25515122 0.02355797
5 0.41135488 0.28909574
key_workerGovernment defined key worker anxiety_disordersAnxiety disorder
1 -0.124341183 -0.044272612
2 0.060472006 -0.454657977
4 0.101148767 -0.002542874
5 0.001245017 -0.223412441
depressive_disordersDepressive disorder eating_disordersEating disorder
1 0.1902082 0.02170131
2 -0.9250352 -0.35626486
4 -0.5760062 -0.08783083
5 0.1813459 0.11619524
obsessive_compulsive_disordersObsessive compulsive disorder
1 -0.49361826
2 -0.01812124
4 -0.08471810
5 -0.71054890
psychotic_disordersPsychotic disorder
1 -1.1089487
2 0.2979795
4 1.0585645
5 0.3955389
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 -0.10297512 0.4668157
2 -0.63152437 -0.4062939
4 -0.57089686 -0.4158154
5 -0.06450619 0.1346350
Std. Errors:
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 0.22811500 0.6032110 0.26430404 0.17427128 0.18209485 0.20977210 0.5336435
2 0.07236235 0.2520531 0.09886115 0.06934149 0.06542848 0.06939268 0.1287891
4 0.21686631 0.7315828 0.23115207 0.17747905 0.16413489 0.19207318 0.5301621
5 0.23077704 0.4167793 0.23100012 0.18229206 0.17175203 0.19401628 0.6060532
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 0.7363648 31.1188540 0.1619219 0.6082385
2 0.1702998 0.3316927 0.0468769 0.2313392
4 0.6085057 30.7023991 0.1574121 92.8201235
5 0.7327581 0.6590994 0.1671625 0.4921610
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
1 0.4604446 0.5948947 1.0195040
2 0.1608439 0.1915771 0.3384476
4 0.3929381 0.4669255 1.0197285
5 0.4604956 0.4301512 41.8687034
ethnicityArab ethnicityOther employment_prior_covidRetired
1 17.15678 0.7271190 1.0324222
2 25.91137 0.3465126 0.2398563
4 20.93417 49.9348290 1.0302115
5 18.21465 0.7273932 1.0453848
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.7558608 0.2643839
2 0.7501030 0.1118624
4 32.9537476 0.3271549
5 0.7564152 0.2641465
employment_changeDecreased employment employment_changeFurloughed
1 0.17912629 0.18834924
2 0.06405068 0.06121457
4 0.15679671 0.15822370
5 0.15705322 0.16594238
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.30415164 0.13619318
2 0.09683825 0.04581928
4 0.25971806 0.12366495
5 0.25016794 0.13069846
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 0.1604032 0.16814462
2 0.0556844 0.05358782
4 0.1462790 0.14340352
5 0.1515374 0.15768240
eating_disordersEating disorder
1 0.2164125
2 0.1009938
4 0.2186787
5 0.1993455
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.26337714
2 0.09774627
4 0.23177739
5 0.26986582
psychotic_disordersPsychotic disorder
1 0.7303503
2 0.2011655
4 0.3593871
5 0.3904275
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.3060431 0.2834739
2 0.1721302 0.1649351
4 0.3802417 0.3883717
5 0.2896535 0.2976452
Residual Deviance: 21886.14
AIC: 22126.14 create z and p values
make Z
z_clustered.masq <- summary(masq.model_clustered)$coefficients/summary(masq.model_clustered)$standard.errors
z_as.is.masq <- summary(masq.model)$coefficients/summary(masq.model)$standard.errorsprint z clusted ethnicity model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 -14.712445 -0.3717288 -1.0221580 -0.2760355 -2.4452911 -1.981547 -1.0050325
2 2.362587 -1.2205919 0.8071782 -2.0903119 -3.8540186 -2.020670 0.7061521
4 -15.721502 -0.9825227 0.5127019 -0.6906407 -0.1690149 -1.290714 -1.3395380
5 -16.112182 0.8492666 -0.3633821 -1.8793933 -1.9384915 -1.349470 -1.4419439
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 -0.7745697 -0.2415868 0.9699317 -0.1129587
2 -0.1802433 1.7160142 -8.7825768 -1.2908465
4 -0.6066670 -0.2405251 3.5204971 -0.1637853
5 -0.9198796 2.6157630 2.6591341 1.0990044
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.2872101 -0.5980897
2 1.9512819 -1.9421174
4 0.6003481 -0.6319982
5 0.7357340 -0.6998069
employment_prior_covidStudent employment_prior_covidUnemployed
1 1.0972313 -1.908343
2 -1.0570011 -7.271362
4 -0.2367257 -2.664270
5 1.2510895 -1.437856
employment_changeDecreased employment employment_changeFurloughed
1 -0.123070 -0.8575631
2 -2.329823 0.2229575
4 1.843026 1.5717644
5 3.536481 2.4301045
employment_changeIncreased employment key_workerGovernment defined key worker
1 -0.95000616 -0.89776690
2 0.55022491 1.32054326
4 0.04637887 0.80620465
5 1.10455385 0.04301414
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 -0.28389025 1.139782
2 -8.20278652 -17.258634
4 -0.07246849 -4.025391
5 -1.53714224 1.152278
eating_disordersEating disorder
1 0.1026512
2 -3.5398664
4 -0.4426668
5 0.5714587
obsessive_compulsive_disordersObsessive compulsive disorder
1 -1.8702644
2 -0.1684613
4 -0.3337714
5 -2.6314898
psychotic_disordersPsychotic disorder
1 -1.513884
2 1.495407
4 2.961968
5 1.014381
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 -0.3368433 1.6420758
2 -3.6671259 -2.4822820
4 -1.4893134 -1.0891670
5 -0.2112421 0.4329931print z ethnicity as is
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 -14.699980 -0.3624156 -1.0154195 -0.2768785 -2.4395091 -1.976360 -1.0021899
2 2.340867 -1.2117310 0.8301373 -2.0918993 -3.8380139 -2.009296 0.7139727
4 -15.750506 -0.9978401 0.5085884 -0.7028762 -0.1558453 -1.275795 -1.3315055
5 -16.153155 0.8100522 -0.3683099 -1.9072018 -1.9151906 -1.309292 -1.4564451
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 -0.7754205 -0.2062436 0.9822266 -0.09843081
2 -0.1759155 1.7243190 -8.7751168 -1.29923928
4 -0.5949832 -0.2059344 3.5249038 -0.10636222
5 -0.9140912 2.6109707 2.6909784 1.11837290
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
1 0.02540575 0.2477286 0.001834269
2 1.58111029 1.3825831 0.441352696
4 0.75588898 1.1013508 -0.233225567
5 -0.05867014 2.0508415 -0.181580899
ethnicityArab ethnicityOther employment_prior_covidRetired
1 -0.1666426 0.4192349 -0.5968306
2 -0.2325253 0.1662093 -1.9571874
4 -0.1879468 -0.1632125 -0.6341390
5 -0.1756347 0.2940247 -0.6962835
employment_prior_covidStudent employment_prior_covidUnemployed
1 1.0938195 -1.920315
2 -1.0588626 -7.276525
4 -0.1993043 -2.640260
5 1.2631821 -1.443192
employment_changeDecreased employment employment_changeFurloughed
1 -0.1390095 -0.8689330
2 -2.3069968 0.2268331
4 1.8846099 1.6125979
5 3.5541444 2.4789019
employment_changeIncreased employment key_workerGovernment defined key worker
1 -0.96081126 -0.91297661
2 0.56389739 1.31979383
4 0.09070594 0.81792592
5 1.15560668 0.00952587
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 -0.27600821 1.131218
2 -8.16490740 -17.262043
4 -0.01738372 -4.016681
5 -1.47430578 1.150071
eating_disordersEating disorder
1 0.1002775
2 -3.5275910
4 -0.4016433
5 0.5828837
obsessive_compulsive_disordersObsessive compulsive disorder
1 -1.8741879
2 -0.1853906
4 -0.3655149
5 -2.6329711
psychotic_disordersPsychotic disorder
1 -1.518379
2 1.481265
4 2.945471
5 1.013092
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 -0.3364726 1.6467679
2 -3.6688764 -2.4633563
4 -1.5014052 -1.0706633
5 -0.2227013 0.4523339two tailed z test p values
p_clustered.masq <- (1 - pnorm(abs(z_clustered.masq), 0, 1)) * 2
p_as.is.masq <- (1 - pnorm(abs(z_as.is.masq), 0, 1)) * 2show p for two tailed z test for ethnicity clustered model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 0.0000000 0.7100948 0.3067061 0.78252084 0.0144735250 0.04752991 0.3148812
2 0.0181479 0.2222406 0.4195638 0.03658980 0.0001161948 0.04331396 0.4800935
4 0.0000000 0.3258424 0.6081598 0.48979138 0.8657849058 0.19680274 0.1803956
5 0.0000000 0.3957330 0.7163195 0.06019081 0.0525632835 0.17718596 0.1493182
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 0.4385940 0.809100347 0.3320805169 0.9100633
2 0.8569615 0.086159441 0.0000000000 0.1967569
4 0.5440719 0.809923197 0.0004307388 0.8699002
5 0.3576357 0.008902833 0.0078341765 0.2717662
ethnicity_clusteredMinoritised ethnic community employment_prior_covidRetired
1 0.77395146 0.54978008
2 0.05102352 0.05212289
4 0.54827429 0.52738803
5 0.46189260 0.48404792
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.2725403 5.634685e-02
2 0.2905111 3.559375e-13
4 0.8128696 7.715559e-03
5 0.2109018 1.504748e-01
employment_changeDecreased employment employment_changeFurloughed
1 0.9020516974 0.39113374
2 0.0198155077 0.82356860
4 0.0653252414 0.11600520
5 0.0004054955 0.01509447
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.3421091 0.3693098
2 0.5821651 0.1866537
4 0.9630083 0.4201249
5 0.2693530 0.9656903
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 7.764945e-01 2.543771e-01
2 2.220446e-16 0.000000e+00
4 9.422291e-01 5.688071e-05
5 1.242585e-01 2.492069e-01
eating_disordersEating disorder
1 0.9182398236
2 0.0004003297
4 0.6580067871
5 0.5676887672
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.061447109
2 0.866220416
4 0.738552065
5 0.008501141
psychotic_disordersPsychotic disorder
1 0.130055305
2 0.134808239
4 0.003056793
5 0.310401087
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.7362350333 0.10057429
2 0.0002452921 0.01305439
4 0.1364048624 0.27608025
5 0.8326983664 0.66501979show p for two tailed z test for ethnicity as is model
(Intercept) age16-18 age19-25 age36-45 age46-55 age56-65 age66-70
1 0.00000000 0.7170415 0.3099059 0.78187339 0.0147072314 0.04811399 0.3162519
2 0.01923902 0.2256154 0.4064612 0.03644753 0.0001240335 0.04450577 0.4752440
4 0.00000000 0.3183569 0.6110408 0.48213290 0.8761549532 0.20202808 0.1830227
5 0.00000000 0.4179102 0.7126422 0.05649446 0.0554681961 0.19043538 0.1452696
age71-75 age76+ genderFemale genderNon-binary/Prefer to self define
1 0.4380912 0.836600671 0.3259882136 0.9215902
2 0.8603603 0.084650277 0.0000000000 0.1938618
4 0.5518547 0.836842135 0.0004236366 0.9152950
5 0.3606689 0.009028562 0.0071242804 0.2634078
ethnicityMixed or multiple ethnicity ethnicityAsian ethnicityBlack
1 0.9797313 0.80434436 0.9985365
2 0.1138528 0.16679272 0.6589577
4 0.4497158 0.27074399 0.8155863
5 0.9532148 0.04028238 0.8559116
ethnicityArab ethnicityOther employment_prior_covidRetired
1 0.8676512 0.6750444 0.55062046
2 0.8161300 0.8679922 0.05032544
4 0.8509184 0.8703511 0.52599011
5 0.8605809 0.7687390 0.48625130
employment_prior_covidStudent employment_prior_covidUnemployed
1 0.2740341 5.481815e-02
2 0.2896624 3.426148e-13
4 0.8420247 8.284237e-03
5 0.2065237 1.489663e-01
employment_changeDecreased employment employment_changeFurloughed
1 0.889442647 0.38488379
2 0.021054999 0.82055350
4 0.059482524 0.10683189
5 0.000379211 0.01317875
employment_changeIncreased employment key_workerGovernment defined key worker
1 0.3366471 0.3612548
2 0.5728240 0.1869039
4 0.9277263 0.4133995
5 0.2478421 0.9923996
anxiety_disordersAnxiety disorder depressive_disordersDepressive disorder
1 7.825418e-01 2.579633e-01
2 2.220446e-16 0.000000e+00
4 9.861305e-01 5.902345e-05
5 1.403993e-01 2.501147e-01
eating_disordersEating disorder
1 0.9201239860
2 0.0004193595
4 0.6879465950
5 0.5599716010
obsessive_compulsive_disordersObsessive compulsive disorder
1 0.060904525
2 0.852922650
4 0.714727063
5 0.008464156
psychotic_disordersPsychotic disorder
1 0.128918841
2 0.138535850
4 0.003224632
5 0.311016405
personality_disordersPersonality disorder bipolar_disordersBipolar disorder
1 0.7365145044 0.09960576
2 0.0002436188 0.01376431
4 0.1332507900 0.28432086
5 0.8237680106 0.65102845Relative risk
Relative risk is the ratio of the probability of choosing one outcome category over the probability of choosing the baseline category.
Relative risk of 1 means there is no difference between outcome and baseline. Greater than 1 indicates outcome is more likely than baseline. Less than one indicates it is less likely than baseline.
RR 1.4 would mean there is a 1.4% chance of outcome occurring relative to baseline etc
if something is below one (e.g. 0.6 x the risk) you can also interpret this as 40% less probability of outcomes (1-0.6 = 0.4). So, if women are 0.6 x as likely to belong to a class as men, that means that there is a 40% lower probability of being in that category for women.
exponentiate the model coefficient to obtain relative risk for all predictors. Can see equations here
clustered ethnicity
relative_risk_clustered.masq <- data.frame(exp(coef(masq.model_clustered))) %>%
rownames_to_column() %>%
rename(class = "rowname",
intercept = "X.Intercept.") %>%
pivot_longer(!class) %>%
rename(predictor = name,
RelativeRisk = value)
relative_risk_clustered.masq$predictor <- gsub("76.","76+",relative_risk_clustered.masq$predictor)
relative_risk_clustered.masq ethnicity
relative_risk_as.is.masq <-data.frame(exp(coef(masq.model))) %>%
rownames_to_column() %>%
rename(class = "rowname",
intercept = "X.Intercept.") %>%
pivot_longer(!class) %>%
rename(predictor = name,
RelativeRisk = value)
relative_risk_as.is.masq$predictor <- gsub("76.","76+",relative_risk_as.is.masq$predictor)
head(relative_risk_as.is.masq)confidence intervals (rr scale)
The confint package has a specific method that is run for multinom objects (see methods(confint)) Obtain the exponent of upper and lower to get relative risk version
First create the CI (converted to relative risk) and reformat it so it is ready to merge with relative risk data for plotting
clustered ethnicity
CI_clustered.masq <- data.frame(exp(confint(masq.model_clustered,level=0.95))) %>%
rownames_to_column()
names(CI_clustered.masq) <- c("predictor",
"1_lower","1_upper",
"2_lower","2_upper",
"4_lower","4_upper",
"5_lower","5_upper"
)
CI_clustered.masq <- CI_clustered.masq %>%
pivot_longer(!predictor) %>%
separate(name, c("class","CI")) %>%
pivot_wider(names_from = CI,
values_from= value)
CI_clustered.masq$predictor <- gsub("-",".",CI_clustered.masq$predictor)
CI_clustered.masq$predictor <- gsub(" ",".",CI_clustered.masq$predictor)
CI_clustered.masq$predictor <- gsub("/",".",CI_clustered.masq$predictor)
CI_clustered.masq$predictor <- gsub("\\(Intercept\\)","intercept",CI_clustered.masq$predictor)
head(CI_clustered.masq)ethnicity as is
CI_as.is.masq<- data.frame(exp(confint(masq.model,level=0.95))) %>%
rownames_to_column()
names(CI_as.is.masq) <- c("predictor",
"1_lower","1_upper",
"2_lower","2_upper",
"4_lower","4_upper",
"5_lower","5_upper"
)
CI_as.is.masq <- CI_as.is.masq %>%
pivot_longer(!predictor) %>%
separate(name, c("class","CI")) %>%
pivot_wider(names_from = CI,
values_from= value)
CI_as.is.masq$predictor <- gsub("-",".",CI_as.is.masq$predictor)
CI_as.is.masq$predictor <- gsub(" ",".",CI_as.is.masq$predictor)
CI_as.is.masq$predictor <- gsub("/",".",CI_as.is.masq$predictor)
CI_as.is.masq$predictor <- gsub("\\(Intercept\\)","intercept",CI_as.is.masq$predictor)
head(CI_as.is.masq)forest plots of relative risk
MASQ
prepare and format plot data
plot_masq_clustered_dat$category <-plot_masq_clustered_dat$predictor
plot_masq_clustered_dat <-
plot_masq_clustered_dat %>%
mutate(category = case_when(grepl("age",predictor) ~ "Age",
grepl("gender",predictor) ~ "Gender",
grepl("ethnicity_clustered",predictor) ~ "Ethnicity",
grepl("employment_prior_covid",predictor) ~ "Employment before covid",
grepl("employment_change",predictor) ~ "Change in employment due to covid",
grepl("key_worker",predictor) ~ "Key worker status",
grepl("anxiety_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("depressive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("eating_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("obsessive_compulsive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("psychotic_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("personality_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("bipolar_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("mental_health_controls",predictor) ~ "Pre-existing mental health diagnosis",
grepl("intercept",predictor) ~ "Intercept"
),
predictor = case_when(grepl("age",predictor) ~ gsub("age","",predictor),
grepl("gender",predictor) ~ gsub("gender","",predictor),
grepl("ethnicity_clustered",predictor) ~ gsub("ethnicity_clustered","",predictor),
grepl("employment_prior_covid",predictor) ~ gsub("employment_prior_covid","",predictor),
grepl("employment_change",predictor) ~ gsub("employment_change","",predictor),
grepl("key_worker",predictor) ~ gsub("key_worker","",predictor),
grepl("anxiety_disorders",predictor) ~ gsub("anxiety_disorders","",predictor),
grepl("depressive_disorders",predictor) ~ gsub("depressive_disorders","",predictor),
grepl("eating_disorders",predictor) ~ gsub("eating_disorders","",predictor),
grepl("obsessive_compulsive_disorders",predictor) ~ gsub("obsessive_compulsive_disorders","",predictor),
grepl("psychotic_disorders",predictor) ~ gsub("psychotic_disorders","",predictor),
grepl("personality_disorders",predictor) ~ gsub("personality_disorders","",predictor),
grepl("bipolar_disorders",predictor) ~ gsub("bipolar_disorders","",predictor),
grepl("mental_health_controls",predictor) ~ gsub("mental_health_controls","",predictor),
grepl("intercept",predictor) ~ "intercept")
)
plot_masq_as.is_dat$category <-plot_masq_as.is_dat$predictor
plot_masq_as.is_dat <-
plot_masq_as.is_dat %>%
mutate(category = case_when(grepl("age",predictor) ~ "Age",
grepl("gender",predictor) ~ "Gender",
grepl("ethnicity",predictor) ~ "Ethnicity",
grepl("employment_prior_covid",predictor) ~ "Employment before covid",
grepl("employment_change",predictor) ~ "Change in employment due to covid",
grepl("key_worker",predictor) ~ "Key worker status",
grepl("anxiety_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("depressive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("eating_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("obsessive_compulsive_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("psychotic_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("personality_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("bipolar_disorders",predictor) ~ "Pre-existing mental health diagnosis",
grepl("mental_health_controls",predictor) ~ "Pre-existing mental health diagnosis",
grepl("intercept",predictor) ~ "Intercept"
),
predictor = case_when(grepl("age",predictor) ~ gsub("age","",predictor),
grepl("gender",predictor) ~ gsub("gender","",predictor),
grepl("ethnicity",predictor) ~ gsub("ethnicity_clustered","",predictor),
grepl("employment_prior_covid",predictor) ~ gsub("employment_prior_covid","",predictor),
grepl("employment_change",predictor) ~ gsub("employment_change","",predictor),
grepl("key_worker",predictor) ~ gsub("key_worker","",predictor),
grepl("anxiety_disorders",predictor) ~ gsub("anxiety_disorders","",predictor),
grepl("depressive_disorders",predictor) ~ gsub("depressive_disorders","",predictor),
grepl("eating_disorders",predictor) ~ gsub("eating_disorders","",predictor),
grepl("obsessive_compulsive_disorders",predictor) ~ gsub("obsessive_compulsive_disorders","",predictor),
grepl("psychotic_disorders",predictor) ~ gsub("psychotic_disorders","",predictor),
grepl("personality_disorders",predictor) ~ gsub("personality_disorders","",predictor),
grepl("bipolar_disorders",predictor) ~ gsub("bipolar_disorders","",predictor),
grepl("mental_health_controls",predictor) ~ gsub("mental_health_controls","",predictor),
grepl("intercept",predictor) ~ "intercept")
)sort variables so categories are together
plot_masq_clustered_dat$category <- factor(plot_masq_clustered_dat$category,
levels=c("Intercept","Gender","Ethnicity","Age","Employment before covid","Change in employment due to covid","Key worker status","Pre-existing mental health diagnosis"))
plot_masq_clustered_dat$predictor <- fct_reorder2(plot_masq_clustered_dat$predictor, plot_masq_clustered_dat$class, plot_masq_clustered_dat$category)
plot_masq_as.is_dat$category <- factor(plot_masq_as.is_dat$category,
levels=c("Intercept","Gender","Ethnicity","Age","Employment before covid","Change in employment due to covid","Key worker status","Pre-existing mental health diagnosis"))
plot_masq_as.is_dat$predictor <- fct_reorder2(plot_masq_as.is_dat$predictor, plot_masq_as.is_dat$class, plot_masq_as.is_dat$category)forest plot clustered ethnicity
Note: smaller category error bars extend beyond a reasonable axis. Wil need fixing for publication.
masq_eth_clustered_forest <-relrisk.forest(plot_masq_clustered_dat,
plot_title="Relative risk (95% CI) of class membership: anhedonia",
sub_title="ethnicity clustered",
lowlim = 0.001,
uplim = 10)
masq_eth_clustered_forestWarning: Removed 1 rows containing missing values (geom_point).
forest plot ethnicity as is
Note: smaller category error bars extend beyond a reasonable axis. Wil need fixing for publication.
masq_eth_as.is_forest <- relrisk.forest(plot_masq_as.is_dat,
plot_title="Relative risk (95% CI) of class membership: anhedonia",
sub_title="ethnicity as is",
lowlim = 0.001,
uplim = 10)
masq_eth_as.is_forestWarning: Removed 3 rows containing missing values (geom_point).
#### save plots
ggsave(
filename=file.path(dirname(dirname(getwd())),"output/ForestPlots/MASQ/anhedonia_RelativeRisk_forestPlot_ethnicityClustered.png"),
plot=masq_eth_clustered_forest,
width = 18, height = 10, dpi = 150, units = "in")Warning: Removed 1 rows containing missing values (geom_point).ggsave(
filename=file.path(dirname(dirname(getwd())),"output/ForestPlots/MASQ/anhedonia_RelativeRisk_forestPlot.png"),
plot=masq_eth_as.is_forest,
width = 18, height = 10, dpi = 150, units = "in")Warning: Removed 3 rows containing missing values (geom_point).